{ "packId": "ai-for-everyone", "packName": "AI for Everyone", "packVersion": "0.1.1", "icon": "✨", "shortName": "AI for Everyone", "description": "A practical, tool-agnostic course for using generative AI well: better prompts, examples, agents, media, documents, research, data, code, verification, privacy, and reusable workflows.", "author": "Flash Feed", "language": "en", "tagsVocabulary": [ "agent-failures", "agent-loop", "agents", "ai", "approval-checkpoints", "aspect-ratio", "aspect-ratios", "audio-workflows", "choose-tools", "coding-cycle", "comparison", "data-code", "decision-support", "decompose", "define-success", "definition", "document-qa", "draft-edit", "evaluate", "everyday-ai", "few-shot", "formulas-charts", "give-context", "handoffs", "high-stakes", "image-prompts", "iterate", "key-concept", "learn-ai", "learn-by-showing", "manipulation", "multimodal", "name-the-task", "numeric", "one-shot", "personal-workflow", "practical-ai", "prepare-data", "privacy", "procedure", "prompt-anatomy", "prompting", "quiz", "research-learning", "research-plan", "role-prompts", "roles-reasoning", "set-constraints", "summarize", "surface-uncertainty", "templates", "test-secure", "thinking-moves", "tools-permissions", "transform", "tutor", "verification", "verify-claims", "verify-protect", "video-planning", "visible-reasons", "vocabulary", "writing-documents", "zero-shot" ], "items": [ { "id": "afe-f-name-the-task-1", "shape": "fact", "title": "Start with an Action", "body": "A strong prompt states what the system should do: draft, compare, classify, extract, critique, plan, or explain. A verb makes the requested behavior easier to evaluate.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Start with an Action is a practical skill you can inspect and improve.", "illustration": { "kind": "photo", "imageSearchTerm": "clear task checklist notebook", "imagePrompt": "Editorial documentary photograph illustrating clear task checklist notebook. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A notebook beside a concise task checklist", "credit": "Pexels · Close-up of a hand writing in a notebook with a checklist for effective task management.", "creditUrl": "https://www.pexels.com/photo/writing-in-notebook-with-checklist-for-task-management-35719566/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-name-the-task-1.webp" }, "uid": "k56wpr56ccf1" }, { "id": "afe-f-name-the-task-2", "shape": "fact", "title": "Name the Deliverable", "body": "The same topic can produce a memo, table, checklist, lesson, or set of questions. Naming the deliverable prevents the system from choosing the container for you.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "practical-ai" ], "illustration": { "imageSearchTerm": "document formats comparison desk", "imagePrompt": "An overhead view of a desk with several distinct blank document layouts, a memo, a table, and a checklist, spread out side by side, with a hand pointing to one.", "alt": "Name the Deliverable", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-name-the-task-2.webp" }, "uid": "6s6lep1fwstvn" }, { "id": "afe-f-name-the-task-3", "shape": "fact", "title": "Separate Goal from Topic", "body": "“Retirement” is a topic. “Explain three retirement-account differences to a new employee” is a goal. A goal combines subject, action, and intended result.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "practical-ai" ], "illustration": { "imageSearchTerm": "compass single destination map", "imagePrompt": "A hand adjusting a compass on an open map so its needle settles on one clearly chosen distant landmark.", "alt": "Separate Goal from Topic", "depictable": false, "credit": "Pexels · Explore destinations with a compass and passport on a world map. Perfect for adventure planning.", "creditUrl": "https://www.pexels.com/photo/passport-and-compass-on-top-of-maps-7235900/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-name-the-task-3.webp" }, "uid": "n9qcue12xmioe" }, { "id": "afe-p-name-the-task-1", "shape": "pair", "sideA": { "modality": "text", "value": "Task verb", "short": "Task verb" }, "sideB": { "modality": "text", "value": "The action the system should perform", "short": "The action the system should perform" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "vocabulary" ], "uid": "1g5foag10hxzu8" }, { "id": "afe-p-name-the-task-2", "shape": "pair", "sideA": { "modality": "text", "value": "Deliverable", "short": "Deliverable" }, "sideB": { "modality": "text", "value": "The concrete output you expect", "short": "The concrete output you expect" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "vocabulary" ], "uid": "f1aagxcodsq3" }, { "id": "afe-d-name-the-task", "shape": "definition", "term": { "modality": "text", "value": "Task specification", "short": "Task specification" }, "definition": { "modality": "text", "value": "A concise statement of the action, subject, and intended result.", "short": "A concise statement of the action" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "definition" ], "uid": "9g0ugugnd71e" }, { "id": "afe-q-name-the-task-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which opening is most actionable?" }, "options": [ { "modality": "text", "value": "Tell me about plans", "short": "Tell me about plans" }, { "modality": "text", "value": "Plans are interesting", "short": "Plans are interesting" }, { "modality": "text", "value": "Compare these three plans", "short": "Compare three plans" }, { "modality": "text", "value": "What do you think?", "short": "What do you think?" } ], "correctIndex": 2, "explanation": "“Compare” names a visible operation and identifies the supplied choices.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "quiz" ], "uid": "sy4r43uhn6vx" }, { "id": "afe-q-name-the-task-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is missing from “Help with my résumé”?" }, "options": [ { "modality": "text", "value": "A model nickname", "short": "A model nickname" }, { "modality": "text", "value": "A longer greeting", "short": "A longer greeting" }, { "modality": "text", "value": "A creative persona", "short": "A creative persona" }, { "modality": "text", "value": "A specific outcome", "short": "A specific outcome" } ], "correctIndex": 3, "explanation": "The request does not say whether to draft, edit, tailor, or critique.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "quiz" ], "uid": "enpve2mmmle" }, { "id": "afe-cmp-name-the-task", "shape": "comparison", "prompt": "Which request gives the system a testable job?", "correct": { "caption": "Draft a five-item checklist for a first apartment.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-name-the-task-draft-five-item.svg" }, "incorrect": { "caption": "Help me get organized for an apartment move sometime soon.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-name-the-task-help-me-moving.svg" }, "explanation": "Naming an action, a scope, and a deliverable — draft, five items, a checklist — gives the system something it can be judged against. \"Help me get organized for an apartment move sometime soon\" names none of the three.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "name-the-task", "comparison" ], "uid": "1qt5lsievr9da" }, { "id": "afe-f-give-context-1", "shape": "fact", "title": "Context Changes the Answer", "body": "Audience, purpose, constraints, and source material determine what a useful answer looks like. The same task can need different content in a classroom, clinic, or kitchen.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Context Changes the Answer is a practical skill you can inspect and improve.", "illustration": { "kind": "photo", "imageSearchTerm": "team briefing audience context", "imagePrompt": "Editorial documentary photograph illustrating team briefing audience context. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Colleagues reviewing a project brief together", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Hideki_Noda_2009_1000km_of_Okayama.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-give-context-1.webp" }, "uid": "16dmom01ispj20" }, { "id": "afe-f-give-context-2", "shape": "fact", "title": "Relevant Beats Exhaustive", "body": "Context should earn its place by changing the output. A focused brief is usually easier to use and verify than a long dump of unrelated history.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "practical-ai" ], "illustration": { "imageSearchTerm": "concise brief tidy desk", "imagePrompt": "A single neat one-page brief centered on a clean desk, set apart from a chaotic overflowing stack of unrelated papers pushed to the side.", "alt": "Relevant Beats Exhaustive", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-give-context-2.webp" }, "uid": "1kxcwtx1shp7g7" }, { "id": "afe-f-give-context-3", "shape": "fact", "title": "Label Inputs Clearly", "body": "Headings such as BACKGROUND, SOURCE TEXT, and REQUIREMENTS help separate evidence from instructions and make complex prompts easier to inspect.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "practical-ai" ], "illustration": { "imageSearchTerm": "organized document sections dividers", "imagePrompt": "A neatly organized document with clearly separated colored section dividers on a desk, viewed from above, with no visible text.", "alt": "Label Inputs Clearly", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:French_Creek_CCC.JPG", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-give-context-3.webp" }, "uid": "cafkib189acm1" }, { "id": "afe-p-give-context-1", "shape": "pair", "sideA": { "modality": "text", "value": "Audience context", "short": "Audience context" }, "sideB": { "modality": "text", "value": "Who will use or read the output", "short": "Who will use or read the output" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "vocabulary" ], "uid": "10i21jy18kf38q" }, { "id": "afe-p-give-context-2", "shape": "pair", "sideA": { "modality": "text", "value": "Situational context", "short": "Situational context" }, "sideB": { "modality": "text", "value": "The circumstances that shape the task", "short": "The circumstances that shape the task" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "vocabulary" ], "uid": "183c87uhvb6pa" }, { "id": "afe-d-give-context", "shape": "definition", "term": { "modality": "text", "value": "Context", "short": "Context" }, "definition": { "modality": "text", "value": "Information supplied with a request that helps determine a relevant response.", "short": "Information supplied with a request that helps determine a relevant response." }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "definition" ], "uid": "zdplur1fx16dd" }, { "id": "afe-q-give-context-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which detail most changes a safety handout?" }, "options": [ { "modality": "text", "value": "The prompt says please", "short": "The prompt says please" }, { "modality": "text", "value": "The font is your favorite", "short": "Favorite font choice" }, { "modality": "text", "value": "The model has a logo", "short": "The model has a logo" }, { "modality": "text", "value": "The readers are children", "short": "Readers are children" } ], "correctIndex": 3, "explanation": "Age changes vocabulary, examples, and safety framing.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "quiz" ], "uid": "1fxcv5b1akzmsd" }, { "id": "afe-q-give-context-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is the best context rule?" }, "options": [ { "modality": "text", "value": "Include details that change the answer", "short": "Answer-changing info" }, { "modality": "text", "value": "Hide uncertainty", "short": "Hide uncertainty" }, { "modality": "text", "value": "Paste every related document", "short": "Paste all related docs" }, { "modality": "text", "value": "Repeat the task five times", "short": "Repeat task five times" } ], "correctIndex": 0, "explanation": "Useful context is selected for relevance, not sheer volume.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "quiz" ], "uid": "1ihu9klcngp87" }, { "id": "afe-cmp-give-context", "shape": "comparison", "prompt": "Which context packet is easier to use?", "correct": { "caption": "A labeled brief with audience, purpose, and source text.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-give-context-labeled-brief-audience.svg" }, "incorrect": { "caption": "An unlabeled dump of every note, memo, and email combined.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-give-context-giant-unlabeled-paste.svg" }, "explanation": "Selected and labeled context makes important evidence easier to find.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "give-context", "comparison" ], "uid": "4rjnj34o082x" }, { "id": "afe-f-set-constraints-1", "shape": "fact", "title": "Format Is Part of the Task", "body": "A correct answer can still be unusable if it arrives in the wrong structure. Specify a table, numbered plan, email, schema, or other container when the container matters.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Format Is Part of the Task is a practical skill you can inspect and improve.", "illustration": { "kind": "photo", "imageSearchTerm": "architect blueprint ruler grid", "imagePrompt": "Editorial documentary photograph illustrating editorial layout constraints grid. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A designer arranging content inside a clear layout grid", "credit": "Unsplash · SOHAM BANERJEE · Unsplash License", "creditUrl": "https://unsplash.com/photos/white-printer-paper-with-black-pencil-SAFF_1rWBqE", "subject": "Top-down view of a hand-drawn, dimensioned single-room floor plan on white paper labelled MEASURED DRAWING, lying on a green gridded Allwin cutting mat with two clear set squares at the top and three mechanical pencils at lower right", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-set-constraints-1.webp" }, "uid": "qg1npjmtz44x" }, { "id": "afe-f-set-constraints-2", "shape": "fact", "title": "Constraints Need Priorities", "body": "Length, completeness, tone, and detail can pull in different directions. Ranking constraints tells the system what should win when all cannot be satisfied.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "practical-ai" ], "illustration": { "imageSearchTerm": "balance scale competing priorities", "imagePrompt": "A balance scale with several different small weighted objects on each side, one side tipping slightly lower to show a clear priority.", "alt": "Constraints Need Priorities", "depictable": false, "credit": "Pexels · Detailed view of a classic mechanical balance scale, offering precision in weight measurement.", "creditUrl": "https://www.pexels.com/photo/libra-in-a-doctors-office-16204377/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-set-constraints-2.webp" }, "uid": "1rszwe6b15fpu" }, { "id": "afe-f-set-constraints-3", "shape": "fact", "title": "Positive Directions Are Clearer", "body": "“Use three short sections with concrete examples” provides a target. A long list of prohibitions may define only what the output should not be.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "practical-ai" ], "illustration": { "imageSearchTerm": "Positive Directions Are Clearer", "imagePrompt": "Positive Directions Are Clearer. “Use three short sections with concrete examples” provides a target. A long list of prohibitions may define only what the output should not be.", "alt": "Positive Directions Are Clearer", "credit": "Pexels · Susanne Jutzeler, suju-foto", "creditUrl": "https://www.pexels.com/photo/three-brown-tiles-on-gray-surface-1154775/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-set-constraints-3.webp" }, "uid": "1xcwq7l1ujkemb" }, { "id": "afe-p-set-constraints-1", "shape": "pair", "sideA": { "modality": "text", "value": "Hard constraint", "short": "Hard constraint" }, "sideB": { "modality": "text", "value": "A boundary the output must satisfy", "short": "A boundary the output must satisfy" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "vocabulary" ], "uid": "ikhg3o70yncs" }, { "id": "afe-p-set-constraints-2", "shape": "pair", "sideA": { "modality": "text", "value": "Soft preference", "short": "Soft preference" }, "sideB": { "modality": "text", "value": "A desirable feature that may flex", "short": "A desirable feature that may flex" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "vocabulary" ], "uid": "1qfnqwpaxcd4z" }, { "id": "afe-d-set-constraints", "shape": "definition", "term": { "modality": "text", "value": "Output format", "short": "Output format" }, "definition": { "modality": "text", "value": "The structure or representation in which a response should be delivered.", "short": "The structure or representation in which a response should be delivered." }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "definition" ], "uid": "1anmuxd1bg4cpr" }, { "id": "afe-q-set-constraints-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which instruction defines a format?" }, "options": [ { "modality": "text", "value": "Be intelligent", "short": "Be intelligent" }, { "modality": "text", "value": "Think carefully", "short": "Think carefully" }, { "modality": "text", "value": "Use your knowledge", "short": "Use your knowledge" }, { "modality": "text", "value": "Return a three-column table", "short": "Return 3-column table" } ], "correctIndex": 3, "explanation": "A table with named columns is an observable structure.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "quiz" ], "uid": "jzhjwdddoevj" }, { "id": "afe-q-set-constraints-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "When constraints conflict, what helps most?" }, "options": [ { "modality": "text", "value": "Remove the task", "short": "Remove the task" }, { "modality": "text", "value": "Add more adjectives", "short": "Add more adjectives" }, { "modality": "text", "value": "Rank their importance", "short": "Rank their importance" }, { "modality": "text", "value": "Ask for certainty", "short": "Ask for certainty" } ], "correctIndex": 2, "explanation": "Priorities tell the system which requirement may flex.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "quiz" ], "uid": "qvsovmz1soke" }, { "id": "afe-cmp-set-constraints", "shape": "comparison", "prompt": "Which constraint is easier to test?", "correct": { "caption": "Use 120–150 words and three labeled sections.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-set-constraints-use-120-150.svg" }, "incorrect": { "caption": "Keep it fairly short and nicely organized.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-set-constraints-keep-fairly-short.svg" }, "explanation": "Numeric limits and named sections can be checked directly.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "set-constraints", "comparison" ], "uid": "xpwhv2153z92a" }, { "id": "afe-f-define-success-1", "shape": "fact", "title": "Criteria Make Quality Testable", "body": "Acceptance criteria translate “good” into observable conditions such as required evidence, fields, length, coverage, or prohibited content.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Acceptance criteria translate “good” into observable conditions such as required evidence, fields, length, coverage, or prohibited content.", "illustration": { "kind": "photo", "imageSearchTerm": "person checking clipboard checklist", "imagePrompt": "Editorial documentary photograph illustrating quality review acceptance criteria. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A reviewer checking work against acceptance criteria", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Andrew_Poston.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-define-success-1.webp" }, "uid": "1y7rqfkimnkgo" }, { "id": "afe-f-define-success-2", "shape": "fact", "title": "Examples Carry Tacit Structure", "body": "A small example can reveal naming, granularity, ordering, and tone that would take many abstract instructions to describe.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "practical-ai" ], "illustration": { "imageSearchTerm": "single sample card template", "imagePrompt": "A single small sample card sitting beside a stack of blank cards, implying it serves as the template for the rest, with no legible text.", "alt": "Examples Carry Tacit Structure", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Single_bananas_in_a_bunch.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-define-success-2.webp" }, "uid": "1335lf41paci24" }, { "id": "afe-f-define-success-3", "shape": "fact", "title": "Self-Checks Still Need Review", "body": "A model can compare its answer with a checklist, but it can also miss its own error. Treat the check as a useful pass, not independent verification.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "practical-ai" ], "illustration": { "imageSearchTerm": "checklist review clipboard desk", "imagePrompt": "A hand marking items on a checklist on a clipboard, with a second, differently colored pen nearby suggesting a separate independent review.", "alt": "Self-Checks Still Need Review", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-define-success-3.webp" }, "uid": "1azhg691ffkanb" }, { "id": "afe-p-define-success-1", "shape": "pair", "sideA": { "modality": "text", "value": "Acceptance criterion", "short": "Acceptance criterion" }, "sideB": { "modality": "text", "value": "A condition the result must meet", "short": "A condition the result must meet" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "vocabulary" ], "uid": "x44lgvsyhkud" }, { "id": "afe-p-define-success-2", "shape": "pair", "sideA": { "modality": "text", "value": "Reference example", "short": "Reference example" }, "sideB": { "modality": "text", "value": "A sample that demonstrates desired features", "short": "A sample that demonstrates desired features" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "vocabulary" ], "uid": "w5cbxy1tlv3ma" }, { "id": "afe-d-define-success", "shape": "definition", "term": { "modality": "text", "value": "Rubric", "short": "Rubric" }, "definition": { "modality": "text", "value": "A set of criteria used to judge the quality of an output.", "short": "A set of criteria used to judge the quality of an output." }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "definition" ], "uid": "128x6m5zk3fx3" }, { "id": "afe-q-define-success-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which criterion is directly checkable?" }, "options": [ { "modality": "text", "value": "Sound very smart", "short": "Sound very smart" }, { "modality": "text", "value": "Use excellent judgment", "short": "Use excellent judgment" }, { "modality": "text", "value": "Every claim cites the packet", "short": "All claims cite packet" }, { "modality": "text", "value": "Make it impressive", "short": "Make it impressive" } ], "correctIndex": 2, "explanation": "Citation presence can be inspected against the source packet.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "quiz" ], "uid": "11w6moxmku53b" }, { "id": "afe-q-define-success-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why ask for a self-check?" }, "options": [ { "modality": "text", "value": "To retrain the model", "short": "To retrain the model" }, { "modality": "text", "value": "To expose missed criteria", "short": "Expose missed criteria" }, { "modality": "text", "value": "To remove human review", "short": "To remove human review" }, { "modality": "text", "value": "To guarantee truth", "short": "To guarantee truth" } ], "correctIndex": 1, "explanation": "The check can surface omissions but does not prove correctness.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "quiz" ], "uid": "1deugkk1tj8jeg" }, { "id": "afe-cmp-define-success", "shape": "comparison", "prompt": "Which revision request supplies a target?", "correct": { "caption": "Keep the meaning; cut to 100 words and retain all dates.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-define-success-keep-meaning-cut.svg" }, "incorrect": { "caption": "Make this much better and more professional, whatever that takes.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-define-success-make-much-better.svg" }, "explanation": "A usable target names both the change and what must survive it: cut to 100 words, keep the meaning and every date. \"Much better and more professional, whatever that takes\" names neither, so nothing can be checked.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "prompt-anatomy", "define-success", "comparison" ], "uid": "1l5sk1a1crt542" }, { "id": "afe-f-zero-shot-1", "shape": "fact", "title": "Zero-Shot Means No Demonstration", "body": "A zero-shot prompt asks the model to perform a task without showing a completed input-output example. Instructions and context can still be detailed.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A zero-shot prompt asks the model to perform a task without showing a completed input-output example. Instructions and context can still be detailed.", "illustration": { "kind": "photo", "imageSearchTerm": "white blank index card mockup", "imagePrompt": "Editorial documentary photograph illustrating blank page direct instruction. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A clean page with one clearly marked instruction", "credit": "Unsplash · Andrew Dunstan · Unsplash License", "creditUrl": "https://unsplash.com/photos/white-paper-on-brown-surface-vmtoLazDg_Y", "subject": "Photo of a single blank white sheet of paper, slightly creased down the middle, centred on brown kraft wrapping paper; no text", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-zero-shot-1.webp" }, "uid": "qe4ywz185tsfx" }, { "id": "afe-f-zero-shot-2", "shape": "fact", "title": "Simple Tasks Often Start Directly", "body": "For a familiar transformation such as turning notes into bullets, a clear instruction may be enough to discover whether examples are needed.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "handwritten notes notepad desk", "imagePrompt": "A handwritten page of messy notes on a desk beside a clean notepad, suggesting a simple direct transformation task, no legible text.", "alt": "Simple Tasks Often Start Directly", "depictable": false, "credit": "Pexels · A close-up of sticky notes with a to-do list, a pen, and a wallet on a wooden desk.", "creditUrl": "https://www.pexels.com/photo/desk-organizer-with-sticky-notes-and-pen-33344612/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-zero-shot-2.webp" }, "uid": "195w9b2q52hda" }, { "id": "afe-f-zero-shot-3", "shape": "fact", "title": "The First Output Is a Probe", "body": "A low-risk first attempt reveals ambiguities in the task, missing context, and format drift. Its failures are information for the next prompt.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "single footprint fresh snow", "imagePrompt": "A single tentative footprint in fresh snow, testing the ground ahead before a full path is committed.", "alt": "The First Output Is a Probe", "depictable": false, "credit": "Pexels · Close-up of a single footprint imprinted on pristine, untouched snow, showcasing winter's serene beauty.", "creditUrl": "https://www.pexels.com/photo/photograph-of-a-footprint-on-snow-15039851/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-zero-shot-3.webp" }, "uid": "598irgjg6xp4" }, { "id": "afe-p-zero-shot-1", "shape": "pair", "sideA": { "modality": "text", "value": "Zero-shot prompt", "short": "Zero-shot prompt" }, "sideB": { "modality": "text", "value": "Instructions without a worked example", "short": "Instructions without a worked example" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "vocabulary" ], "uid": "13wz0b21ljaftq" }, { "id": "afe-p-zero-shot-2", "shape": "pair", "sideA": { "modality": "text", "value": "Probe run", "short": "Probe run" }, "sideB": { "modality": "text", "value": "A first attempt used to discover failure modes", "short": "A first attempt used to discover failure modes" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "vocabulary" ], "uid": "rx2oktrlwc07" }, { "id": "afe-d-zero-shot", "shape": "definition", "term": { "modality": "text", "value": "Zero-shot prompting", "short": "Zero-shot prompting" }, "definition": { "modality": "text", "value": "Requesting a task without supplying a completed demonstration.", "short": "Requesting a task without supplying a completed demonstration." }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "definition" ], "uid": "109eeyb1pj77r1" }, { "id": "afe-q-zero-shot-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What can a zero-shot prompt include?" }, "options": [ { "modality": "text", "value": "A trained adapter", "short": "A trained adapter" }, { "modality": "text", "value": "Context and constraints", "short": "Context & constraints" }, { "modality": "text", "value": "No source material", "short": "No source material" }, { "modality": "text", "value": "Only one sentence", "short": "Only one sentence" } ], "correctIndex": 1, "explanation": "The term concerns demonstrations, not the amount of instruction.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "quiz" ], "uid": "nap4yq1h1s6tu" }, { "id": "afe-q-zero-shot-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "When is zero-shot a good start?" }, "options": [ { "modality": "text", "value": "A hidden grading label", "short": "A hidden grading label" }, { "modality": "text", "value": "A novel private code", "short": "A novel private code" }, { "modality": "text", "value": "A safety-critical decision", "short": "Safety-critical choice" }, { "modality": "text", "value": "A familiar low-risk task", "short": "Familiar low-risk task" } ], "correctIndex": 3, "explanation": "Simple low-risk tasks let you test the direct request cheaply.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "quiz" ], "uid": "ydpoo8y859bs" }, { "id": "afe-cmp-zero-shot", "shape": "comparison", "prompt": "Which is a zero-shot request?", "correct": { "caption": "Classify each note as action, decision, or question, no example given.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-zero-shot-classify-note-action.svg" }, "incorrect": { "caption": "Here is one labeled example first; classify the next note the same way.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-zero-shot-here-labeled-example.svg" }, "explanation": "Zero-shot means the instruction stands alone — classify each note into the three categories, with no worked demonstration. Supplying a labeled example first makes it one-shot.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "zero-shot", "comparison" ], "uid": "y71dj81lue9h8" }, { "id": "afe-f-one-shot-1", "shape": "fact", "title": "One Example Can Define a Pattern", "body": "A worked input-output pair can demonstrate the expected labels, level of detail, or transformation more precisely than style adjectives.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A worked input-output pair can demonstrate the expected labels, level of detail, or transformation more precisely than style adjectives.", "illustration": { "kind": "photo", "imageSearchTerm": "two blank index cards comparison", "imagePrompt": "Editorial documentary photograph illustrating before and after example cards. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Two cards showing a simple before-and-after example", "credit": "Pexels · A minimalist mockup featuring two blank business cards on a textured surface.", "creditUrl": "https://www.pexels.com/photo/blank-rectangle-white-paper-on-black-textile-8489955/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-one-shot-1.webp" }, "uid": "170qzb31mwpudx" }, { "id": "afe-f-one-shot-2", "shape": "fact", "title": "Examples Need Boundaries", "body": "Clear markers around EXAMPLE INPUT, EXAMPLE OUTPUT, and YOUR INPUT reduce confusion between the demonstration and the live task.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "two labeled folders desk", "imagePrompt": "Two clearly separated folders side by side on a desk, one smaller marked as a sample and one larger as the main working set, no visible text.", "alt": "Examples Need Boundaries", "depictable": false, "credit": "Pexels · Close-up of two red lever arch files on a wooden desk in a modern office setting.", "creditUrl": "https://www.pexels.com/photo/two-red-paper-organizers-265024/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-one-shot-2.webp" }, "uid": "1ge83af51zqlp" }, { "id": "afe-f-one-shot-3", "shape": "fact", "title": "Choose a Representative Case", "body": "An example should exercise the rule you care about. An unusually easy or exceptional case can teach the wrong boundary.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "representative item row selection", "imagePrompt": "A single typical item highlighted among a diverse row of objects, chosen as the representative example rather than an outlier at either end.", "alt": "Choose a Representative Case", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-one-shot-3.webp" }, "uid": "xbtabn2ag3sp" }, { "id": "afe-p-one-shot-1", "shape": "pair", "sideA": { "modality": "text", "value": "One-shot prompt", "short": "One-shot prompt" }, "sideB": { "modality": "text", "value": "Instructions plus one demonstration", "short": "Instructions plus one demonstration" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "vocabulary" ], "uid": "17ay6le1c0l6li" }, { "id": "afe-p-one-shot-2", "shape": "pair", "sideA": { "modality": "text", "value": "Representative example", "short": "Representative example" }, "sideB": { "modality": "text", "value": "A case that exposes the intended rule", "short": "A case that exposes the intended rule" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "vocabulary" ], "uid": "1p6hcljrf375" }, { "id": "afe-d-one-shot", "shape": "definition", "term": { "modality": "text", "value": "Demonstration", "short": "Demonstration" }, "definition": { "modality": "text", "value": "A worked example included in a prompt to show how an input should map to an output.", "short": "A worked example included in a prompt to show how an input should map to an output." }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "definition" ], "uid": "o818vg1c0qlyk" }, { "id": "afe-q-one-shot-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is one-shot prompting?" }, "options": [ { "modality": "text", "value": "Using one-word answers", "short": "Using one-word answers" }, { "modality": "text", "value": "Giving one worked example", "short": "One worked example" }, { "modality": "text", "value": "Training one parameter", "short": "Training one parameter" }, { "modality": "text", "value": "Running the model once", "short": "Running the model once" } ], "correctIndex": 1, "explanation": "The “shot” is a demonstration, not an execution count.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "quiz" ], "uid": "1w2j2w3hogfz5" }, { "id": "afe-q-one-shot-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why delimit an example?" }, "options": [ { "modality": "text", "value": "To add hidden memory", "short": "To add hidden memory" }, { "modality": "text", "value": "To separate it from live input", "short": "Separate live input" }, { "modality": "text", "value": "To reduce all token use", "short": "Reduce all token use" }, { "modality": "text", "value": "To guarantee accuracy", "short": "To guarantee accuracy" } ], "correctIndex": 1, "explanation": "Boundaries clarify which text is instruction, example, and task.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "quiz" ], "uid": "16vllrcde1pnc" }, { "id": "afe-cmp-one-shot", "shape": "comparison", "prompt": "Which demonstration is easier to transfer?", "correct": { "caption": "A labeled input and output with the desired structure.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-one-shot-labeled-input-output.svg" }, "incorrect": { "caption": "A finished answer with no indication of what it demonstrates.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-one-shot-finished-indication-demonstrates.svg" }, "explanation": "Labeled boundaries reveal the mapping the learner should copy.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "one-shot", "comparison" ], "uid": "1uoy6yn17ug5r9" }, { "id": "afe-f-few-shot-1", "shape": "fact", "title": "Variation Teaches Boundaries", "body": "Several examples can show a category's center, edge, and near miss. Diversity is more informative than repeating nearly identical easy cases.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Several examples can show a category's center, edge, and near miss. Diversity is more informative than repeating nearly identical easy cases.", "illustration": { "kind": "photo", "imageSearchTerm": "index cards varied examples", "imagePrompt": "Editorial documentary photograph illustrating index cards varied examples. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Several index cards showing varied examples", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Gfx14936r.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-few-shot-1.webp" }, "uid": "9oi7f2hj0ani" }, { "id": "afe-f-few-shot-2", "shape": "fact", "title": "Examples Communicate More Than Labels", "body": "Demonstrations can reveal output format, available labels, and the distribution of inputs, not only a direct reasoning rule.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "sample cards fanned table", "imagePrompt": "A small stack of varied sample cards fanned out on a table, each showing a different layout, implying they teach more than a single label could.", "alt": "Examples Communicate More Than Labels", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-few-shot-2.webp" }, "uid": "1cwxgeq12lck26" }, { "id": "afe-f-few-shot-3", "shape": "fact", "title": "Curate the Set", "body": "More examples use more context and can introduce contradictions. Keep examples relevant, accurate, and distinct enough to add information.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "practical-ai" ], "illustration": { "imageSearchTerm": "hand selecting cards pile", "imagePrompt": "A hand selecting a few distinct cards from a much larger scattered pile, setting the rest aside.", "alt": "Curate the Set", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Brother_Can_you_Spare_a_Dime.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-few-shot-3.webp" }, "uid": "uo2j0l1lt1m8b" }, { "id": "afe-p-few-shot-1", "shape": "pair", "sideA": { "modality": "text", "value": "Few-shot prompt", "short": "Few-shot prompt" }, "sideB": { "modality": "text", "value": "Instructions plus several demonstrations", "short": "Instructions plus several demonstrations" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "vocabulary" ], "uid": "jabw7p1yzdop7" }, { "id": "afe-p-few-shot-2", "shape": "pair", "sideA": { "modality": "text", "value": "Boundary example", "short": "Boundary example" }, "sideB": { "modality": "text", "value": "A case near the edge of a category", "short": "A case near the edge of a category" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "vocabulary" ], "uid": "9rw9e01w4zjbc" }, { "id": "afe-d-few-shot", "shape": "definition", "term": { "modality": "text", "value": "Few-shot prompting", "short": "Few-shot prompting" }, "definition": { "modality": "text", "value": "Prompting with a small set of worked demonstrations.", "short": "Prompting with a small set of worked demonstrations." }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "definition" ], "uid": "1sl4q5v1bhi0fl" }, { "id": "afe-q-few-shot-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What makes a few-shot set useful?" }, "options": [ { "modality": "text", "value": "Hidden instructions", "short": "Hidden instructions" }, { "modality": "text", "value": "Varied boundary coverage", "short": "Varied boundary cases" }, { "modality": "text", "value": "Random long passages", "short": "Random long passages" }, { "modality": "text", "value": "Maximum repetition", "short": "Maximum repetition" } ], "correctIndex": 1, "explanation": "Distinct cases teach more of the decision boundary.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "quiz" ], "uid": "5z9n79km0dkr" }, { "id": "afe-q-few-shot-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is a risk of extra examples?" }, "options": [ { "modality": "text", "value": "Automatic retraining", "short": "Automatic retraining" }, { "modality": "text", "value": "Guaranteed copying", "short": "Guaranteed copying" }, { "modality": "text", "value": "Contradictions and clutter", "short": "Conflict and clutter" }, { "modality": "text", "value": "Loss of all context", "short": "Loss of all context" } ], "correctIndex": 2, "explanation": "Poorly curated examples can compete or disagree.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "quiz" ], "uid": "iwgl3t1dro59f" }, { "id": "afe-proc-few-shot", "shape": "procedure", "goal": "Build a few-shot example set", "steps": [ "Name the distinction the examples must teach", "Add one ordinary positive case", "Add a useful edge case", "Add a near miss or contrasting case", "Check labels, format, and instructions for consistency" ], "notes": "Prefer a small, coherent set over a long pile of redundant examples.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "few-shot", "procedure" ], "uid": "hmj9a41npp32s" }, { "id": "afe-f-iterate-1", "shape": "fact", "title": "Prompting Is Test-Driven", "body": "A prompt version is useful when it performs well on representative cases and a stated rubric, not merely when one output feels impressive.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A prompt version is useful when it performs well on representative cases and a stated rubric, not merely when one output feels impressive.", "illustration": { "kind": "photo", "imageSearchTerm": "paper drafts red pen revisions", "imagePrompt": "Editorial documentary photograph illustrating draft revision feedback loop. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Successive drafts arranged in a feedback loop", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Gateway-Tracing-Paper.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-iterate-1.webp" }, "uid": "1iy5v71o7oba7" }, { "id": "afe-f-iterate-2", "shape": "fact", "title": "Change One Major Variable", "body": "Holding the test cases steady while changing one instruction or example makes the effect easier to interpret.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "practical-ai" ], "illustration": { "imageSearchTerm": "test tubes single variable", "imagePrompt": "A row of identical test tubes with only one showing a different colored liquid, representing a single changed variable in an otherwise controlled set.", "alt": "Change One Major Variable", "depictable": false, "credit": "Pexels · Laboratory test tubes in a plastic rack on a reflective blue surface.", "creditUrl": "https://www.pexels.com/photo/a-collection-of-blue-cap-plastic-test-tubes-in-a-rack-7722535/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-iterate-2.webp" }, "uid": "95cjxk272mdc" }, { "id": "afe-f-iterate-3", "shape": "fact", "title": "Diagnose Before Rewriting", "body": "Different failures need different fixes. Missing evidence is not repaired by a more enthusiastic tone, and a wrong format may not need a new model.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "practical-ai" ], "illustration": { "imageSearchTerm": "mechanic diagnostic engine inspection", "imagePrompt": "A mechanic examining an engine closely with a diagnostic tool before reaching for any wrench, representing diagnosis before a fix.", "alt": "Diagnose Before Rewriting", "depictable": false, "credit": "Pexels · An auto mechanic in blue coveralls inspecting a car's engine in a garage.", "creditUrl": "https://www.pexels.com/photo/man-in-blue-coveralls-checking-on-a-vehicle-6870295/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-iterate-3.webp" }, "uid": "1au9jea1xixpua" }, { "id": "afe-p-iterate-1", "shape": "pair", "sideA": { "modality": "text", "value": "Prompt regression", "short": "Prompt regression" }, "sideB": { "modality": "text", "value": "A previously good case fails after a change", "short": "A previously good case fails after a change" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "vocabulary" ], "uid": "1n3y9c2116un6i" }, { "id": "afe-p-iterate-2", "shape": "pair", "sideA": { "modality": "text", "value": "Test set", "short": "Test set" }, "sideB": { "modality": "text", "value": "Representative inputs used across prompt versions", "short": "Representative inputs used across prompt versions" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "vocabulary" ], "uid": "k2hkbok1gll4" }, { "id": "afe-d-iterate", "shape": "definition", "term": { "modality": "text", "value": "Prompt iteration", "short": "Prompt iteration" }, "definition": { "modality": "text", "value": "A cycle of testing, diagnosing, revising, and comparing instructions.", "short": "A cycle of testing" }, "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "definition" ], "uid": "dd6j4k9l9dmo" }, { "id": "afe-q-iterate-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What makes prompt comparison fairer?" }, "options": [ { "modality": "text", "value": "Use the same test cases", "short": "Same test cases" }, { "modality": "text", "value": "Hide the rubric", "short": "Hide the rubric" }, { "modality": "text", "value": "Change every variable", "short": "Change every variable" }, { "modality": "text", "value": "Judge one lucky output", "short": "Judge one lucky output" } ], "correctIndex": 0, "explanation": "Stable cases let you attribute changes more carefully.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "quiz" ], "uid": "owejvd16g1n8v" }, { "id": "afe-q-iterate-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "A response lacks evidence. Best first fix?" }, "options": [ { "modality": "text", "value": "Add a poetic persona", "short": "Add a poetic persona" }, { "modality": "text", "value": "Supply sources and require citations", "short": "Sources plus citations" }, { "modality": "text", "value": "Ask for a longer greeting", "short": "Longer greeting" }, { "modality": "text", "value": "Increase enthusiasm", "short": "Increase enthusiasm" } ], "correctIndex": 1, "explanation": "The repair should address the missing evidence.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "quiz" ], "uid": "vif5we1guwzwq" }, { "id": "afe-proc-iterate", "shape": "procedure", "goal": "Run a prompt improvement cycle", "steps": [ "Save the current prompt and representative inputs", "Score outputs with a short rubric", "Name the dominant failure pattern", "Change one major instruction or example", "Retest and keep the version only if it improves the set" ], "notes": "Record regressions as well as wins; a prompt can improve one case while harming another.", "source": { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, "tags": [ "learn-by-showing", "iterate", "procedure" ], "uid": "wjuqb5184xlqr" }, { "id": "afe-f-role-prompts-1", "shape": "fact", "title": "A Role Sets a Lens", "body": "Roles can cue tone, audience, priorities, or review standards. They are most useful when the desired perspective can be described and evaluated.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Roles can cue tone, audience, priorities, or review standards. They are most useful when the desired perspective can be described and evaluated.", "illustration": { "kind": "photo", "imageSearchTerm": "editor reviewing manuscript perspective", "imagePrompt": "Editorial documentary photograph illustrating editor reviewing manuscript perspective. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "An editor reviewing a manuscript from a professional perspective", "credit": "Pexels · Elderly woman reviewing a manuscript with a typewriter at a wooden desk.", "creditUrl": "https://www.pexels.com/photo/a-woman-reading-and-marking-a-composition-7967597/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-role-prompts-1.webp" }, "uid": "t6h7f31hmf3vp" }, { "id": "afe-f-role-prompts-2", "shape": "fact", "title": "A Persona Is Not a Credential", "body": "“Act as a doctor” does not confer medical training, accountability, or access to a patient's record. The output still needs appropriate evidence and review.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "practical-ai" ], "illustration": { "imageSearchTerm": "empty coat stethoscope chair", "imagePrompt": "A white lab coat and stethoscope draped over an empty chair with no one wearing them, suggesting an unfilled role.", "alt": "A Persona Is Not a Credential", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:2015_Vauxhall_Vivaro,_rear_interior.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-role-prompts-2.webp" }, "uid": "98t7mq5259va" }, { "id": "afe-f-role-prompts-3", "shape": "fact", "title": "Task Details Usually Matter More", "body": "A clear assignment, source packet, and rubric provide stronger control than an impressive job title alone.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "practical-ai" ], "illustration": { "imageSearchTerm": "rubric source packet desk", "imagePrompt": "A detailed rubric and source packet spread open on a desk, positioned more prominently than a small name badge pushed to the side.", "alt": "Task Details Usually Matter More", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-role-prompts-3.webp" }, "uid": "14ie4hea0e2ai" }, { "id": "afe-p-role-prompts-1", "shape": "pair", "sideA": { "modality": "text", "value": "Role prompt", "short": "Role prompt" }, "sideB": { "modality": "text", "value": "An instruction to adopt a perspective or standard", "short": "An instruction to adopt a perspective or standard" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "vocabulary" ], "uid": "ib8kcelccfn6" }, { "id": "afe-p-role-prompts-2", "shape": "pair", "sideA": { "modality": "text", "value": "Expertise claim", "short": "Expertise claim" }, "sideB": { "modality": "text", "value": "A statement of actual competence or qualification", "short": "A statement of actual competence or qualification" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "vocabulary" ], "uid": "1spby7ix5u9c2" }, { "id": "afe-d-role-prompts", "shape": "definition", "term": { "modality": "text", "value": "Persona prompting", "short": "Persona prompting" }, "definition": { "modality": "text", "value": "Directing a model to respond from a named role or perspective.", "short": "Directing a model to respond from a named role or perspective." }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "definition" ], "uid": "8qqadz6oa1vl" }, { "id": "afe-q-role-prompts-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What can a role prompt reliably provide?" }, "options": [ { "modality": "text", "value": "A professional license", "short": "A professional license" }, { "modality": "text", "value": "A requested perspective", "short": "Requested perspective" }, { "modality": "text", "value": "Private case knowledge", "short": "Private case knowledge" }, { "modality": "text", "value": "Guaranteed factual accuracy", "short": "Guaranteed accuracy" } ], "correctIndex": 1, "explanation": "A role is a behavioral cue, not verified expertise.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "quiz" ], "uid": "wtno5h1vbf3u3" }, { "id": "afe-q-role-prompts-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which role instruction is most testable?" }, "options": [ { "modality": "text", "value": "Never make an error", "short": "Never make an error" }, { "modality": "text", "value": "Be the world's best genius", "short": "World's best genius" }, { "modality": "text", "value": "Know every hidden fact", "short": "Know every hidden fact" }, { "modality": "text", "value": "Review this as a copy editor", "short": "Review as copy editor" } ], "correctIndex": 3, "explanation": "Copy editing names a concrete review lens.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "quiz" ], "uid": "yxv4ee7fh9re" }, { "id": "afe-cmp-role-prompts", "shape": "comparison", "prompt": "Which prompt uses a role responsibly?", "correct": { "caption": "As a copy editor, flag ambiguity and cite each sentence.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-role-prompts-copy-editor-flag.svg" }, "incorrect": { "caption": "As a lawyer, guarantee this contract is enforceable everywhere.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-role-prompts-lawyer-guarantee-contract.svg" }, "explanation": "A role used responsibly specifies observable behavior — flag ambiguity, cite each sentence. Casting the model as a lawyer to \"guarantee\" enforceability treats the persona as authority it does not have.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "role-prompts", "comparison" ], "uid": "1k00m0w1sqzbrc" }, { "id": "afe-f-visible-reasons-1", "shape": "fact", "title": "Request Inspectable Artifacts", "body": "Citations, formulas, assumptions, intermediate tables, and test results give a reviewer concrete objects to check.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Citations, formulas, assumptions, intermediate tables, and test results give a reviewer concrete objects to check.", "illustration": { "kind": "photo", "imageSearchTerm": "highlighter research paper notes desk", "imagePrompt": "Editorial documentary photograph illustrating evidence notes decision rationale. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A decision note linking claims to visible evidence", "credit": "Pexels · Hands highlighting and taking notes on a document with pen and highlighter.", "creditUrl": "https://www.pexels.com/photo/close-up-photo-of-a-person-using-a-pink-highlighter-7681064/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-visible-reasons-1.webp" }, "uid": "odtcqi1em5upa" }, { "id": "afe-f-visible-reasons-2", "shape": "fact", "title": "Length Is Not Proof", "body": "A detailed explanation can still be unsupported, post-hoc, or wrong. Fluency and confidence do not substitute for evidence.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "practical-ai" ], "illustration": { "imageSearchTerm": "thick document stack desk", "imagePrompt": "A very thick bound document stacked next to a single small verification stamp, implying volume alone does not equal proof.", "alt": "Length Is Not Proof", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Ode_To_My_Last_Night_In_Porto_(4563947986).jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-visible-reasons-2.webp" }, "uid": "sn51wm1wm457a" }, { "id": "afe-f-visible-reasons-3", "shape": "fact", "title": "Match the Reason to the Task", "body": "A calculation needs shown inputs and arithmetic; a source-based summary needs claim-to-passage links; a recommendation needs criteria and tradeoffs.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "practical-ai" ], "illustration": { "imageSearchTerm": "calculator highlighted passage desk", "imagePrompt": "A desk split between a calculator surrounded by numbers and a book page with one passage highlighted, representing different kinds of justification.", "alt": "Match the Reason to the Task", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-visible-reasons-3.webp" }, "uid": "1kr80dc1iem9nk" }, { "id": "afe-p-visible-reasons-1", "shape": "pair", "sideA": { "modality": "text", "value": "Concise rationale", "short": "Concise rationale" }, "sideB": { "modality": "text", "value": "A short justification tied to evidence", "short": "A short justification tied to evidence" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "vocabulary" ], "uid": "11w235hpv2kv3" }, { "id": "afe-p-visible-reasons-2", "shape": "pair", "sideA": { "modality": "text", "value": "Verifiable trace", "short": "Verifiable trace" }, "sideB": { "modality": "text", "value": "An artifact a reviewer can independently check", "short": "An artifact a reviewer can independently check" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "vocabulary" ], "uid": "zjha1l178mz8z" }, { "id": "afe-d-visible-reasons", "shape": "definition", "term": { "modality": "text", "value": "Rationale", "short": "Rationale" }, "definition": { "modality": "text", "value": "A stated justification for a conclusion or choice.", "short": "A stated justification for a conclusion or choice." }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "definition" ], "uid": "1qiq2bbmh43jp" }, { "id": "afe-q-visible-reasons-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which request is easiest to verify?" }, "options": [ { "modality": "text", "value": "Show inputs and formula", "short": "Show inputs & formula" }, { "modality": "text", "value": "Write a longer answer", "short": "Write a longer answer" }, { "modality": "text", "value": "Be completely certain", "short": "Be completely certain" }, { "modality": "text", "value": "Think more deeply", "short": "Think more deeply" } ], "correctIndex": 0, "explanation": "Inputs and a formula can be recalculated.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "quiz" ], "uid": "1f8oo09phui37" }, { "id": "afe-q-visible-reasons-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What does a long explanation prove?" }, "options": [ { "modality": "text", "value": "That every step is true", "short": "Every step is true" }, { "modality": "text", "value": "That risk is eliminated", "short": "Risk is eliminated" }, { "modality": "text", "value": "Only that text was produced", "short": "Only text was produced" }, { "modality": "text", "value": "That sources were checked", "short": "Sources were checked" } ], "correctIndex": 2, "explanation": "Length alone does not validate reasoning or evidence.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "quiz" ], "uid": "1ub04ypz4i8kz" }, { "id": "afe-cmp-visible-reasons", "shape": "comparison", "prompt": "Which justification is more auditable?", "correct": { "caption": "Recommendation, criteria, evidence, and remaining uncertainty.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-visible-reasons-recommendation-criteria-evidence.svg" }, "incorrect": { "caption": "Trust this answer because the reasoning was extremely detailed.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-visible-reasons-trust-because-reasoning.svg" }, "explanation": "Auditable support connects the conclusion to checkable inputs.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "visible-reasons", "comparison" ], "uid": "vnmbxticdktv" }, { "id": "afe-f-decompose-1", "shape": "fact", "title": "Stages Expose Decisions", "body": "Planning, drafting, and checking as separate phases makes assumptions and intermediate choices easier to inspect.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Planning, drafting, and checking as separate phases makes assumptions and intermediate choices easier to inspect.", "illustration": { "kind": "photo", "imageSearchTerm": "workflow stages sticky notes", "imagePrompt": "Editorial documentary photograph illustrating workflow stages sticky notes. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A complex task divided into clear workflow stages", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:AREMA_Use_cases_diagramm.JPG", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decompose-1.webp" }, "uid": "emz3ok1wppy6w" }, { "id": "afe-f-decompose-2", "shape": "fact", "title": "Approval Can Sit Between Stages", "body": "A human can approve the outline before prose is produced or approve a query before it touches a database.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "practical-ai" ], "illustration": { "imageSearchTerm": "approval stamp document checkpoint", "imagePrompt": "A hand stamping approval on a document resting at a checkpoint between two stacked trays, no visible text.", "alt": "Approval Can Sit Between Stages", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Gepr%C3%BCfte_Servicequalit%C3%A4t.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decompose-2.webp" }, "uid": "bqfkqn1c4ch05" }, { "id": "afe-f-decompose-3", "shape": "fact", "title": "Complexity Has a Cost", "body": "Extra steps add latency and opportunities for drift. Decompose only where a stage creates control or useful evidence.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "practical-ai" ], "illustration": { "imageSearchTerm": "chain interlocking gears mechanism", "imagePrompt": "A long, winding chain of many small interlocking gears connected in sequence, illustrating the added complexity and friction of extra steps.", "alt": "Complexity Has a Cost", "depictable": false, "credit": "Pexels · Detailed view of interlocking metal gears in industrial machinery.", "creditUrl": "https://www.pexels.com/photo/close-up-of-metal-cogs-11518280/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decompose-3.webp" }, "uid": "7y4dn8b6cttw" }, { "id": "afe-p-decompose-1", "shape": "pair", "sideA": { "modality": "text", "value": "Planning stage", "short": "Planning stage" }, "sideB": { "modality": "text", "value": "Chooses an approach before production", "short": "Chooses an approach before production" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "vocabulary" ], "uid": "14n0m5x1fosu7v" }, { "id": "afe-p-decompose-2", "shape": "pair", "sideA": { "modality": "text", "value": "Review stage", "short": "Review stage" }, "sideB": { "modality": "text", "value": "Checks output against requirements and evidence", "short": "Checks output against requirements and evidence" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "vocabulary" ], "uid": "zqdr8a1m31nm6" }, { "id": "afe-d-decompose", "shape": "definition", "term": { "modality": "text", "value": "Task decomposition", "short": "Task decomposition" }, "definition": { "modality": "text", "value": "Dividing a complex job into smaller, inspectable stages.", "short": "Dividing a complex job into smaller" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "definition" ], "uid": "2e4e4wsq0gys" }, { "id": "afe-q-decompose-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why approve an outline first?" }, "options": [ { "modality": "text", "value": "Remove all later review", "short": "Remove later review" }, { "modality": "text", "value": "Retrain the model", "short": "Retrain the model" }, { "modality": "text", "value": "Catch direction errors early", "short": "Catch direction errors" }, { "modality": "text", "value": "Guarantee every fact", "short": "Guarantee every fact" } ], "correctIndex": 2, "explanation": "An early checkpoint prevents expensive drafting on the wrong plan.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "quiz" ], "uid": "74lpld1ynu8yv" }, { "id": "afe-q-decompose-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "When is decomposition least useful?" }, "options": [ { "modality": "text", "value": "A trivial reversible task", "short": "Trivial safe task" }, { "modality": "text", "value": "A multi-source report", "short": "A multi-source report" }, { "modality": "text", "value": "A database update", "short": "A database update" }, { "modality": "text", "value": "A complex migration", "short": "A complex migration" } ], "correctIndex": 0, "explanation": "Extra stages may cost more than they control for simple work.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "quiz" ], "uid": "m4eagchrl2tc" }, { "id": "afe-proc-decompose", "shape": "procedure", "goal": "Decompose a complex AI-assisted task", "steps": [ "List the available inputs and unknowns", "Ask for a plan with decision points", "Approve or revise the plan", "Generate one stage at a time", "Check each deliverable before continuing" ], "notes": "Keep a stage only when it reveals a choice, artifact, or approval boundary.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "decompose", "procedure" ], "uid": "1475j72vnr8ea" }, { "id": "afe-f-surface-uncertainty-1", "shape": "fact", "title": "Assumptions Fill Gaps", "body": "When a request leaves key details open, a model may choose plausible values rather than stop. Making assumptions explicit keeps them reviewable.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "When a request leaves key details open, a model may choose plausible values rather than stop. Making assumptions explicit keeps them reviewable.", "illustration": { "kind": "photo", "imageSearchTerm": "question mark sticky notes planning", "imagePrompt": "Editorial documentary photograph illustrating assumption log uncertainty notes. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A project log separating facts, assumptions, and open questions", "credit": "Pexels · A young boy in a yellow shirt ponders yellow sticky notes on a white wall.", "creditUrl": "https://www.pexels.com/photo/a-boy-in-yellow-shirt-3905693/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-surface-uncertainty-1.webp" }, "uid": "nhsemb1vkfied" }, { "id": "afe-f-surface-uncertainty-2", "shape": "fact", "title": "Unknown Is a Valid Output", "body": "A responsible response can identify missing information, state what evidence would resolve it, or decline to make a high-stakes guess.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "practical-ai" ], "illustration": { "imageSearchTerm": "foggy road signpost fork", "imagePrompt": "A signpost at a fork in a foggy road with one direction left deliberately blank, representing an honest unknown rather than a guess.", "alt": "Unknown Is a Valid Output", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Foggy_landscape.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-surface-uncertainty-2.webp" }, "uid": "1ew391rbi3rtl" }, { "id": "afe-f-surface-uncertainty-3", "shape": "fact", "title": "Confidence Needs Calibration", "body": "A numeric confidence claim is useful only when its relationship to actual accuracy has been tested for the task and setting.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "practical-ai" ], "illustration": { "imageSearchTerm": "gauge dial calibration check", "imagePrompt": "A dial gauge with a needle pointing to a precise reading, being checked side by side against a separate reference gauge for calibration.", "alt": "Confidence Needs Calibration", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-surface-uncertainty-3.webp" }, "uid": "1f6cuutjm256j" }, { "id": "afe-p-surface-uncertainty-1", "shape": "pair", "sideA": { "modality": "text", "value": "Assumption", "short": "Assumption" }, "sideB": { "modality": "text", "value": "A premise used without confirmation", "short": "A premise used without confirmation" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "vocabulary" ], "uid": "1imb0rm1ks9nfu" }, { "id": "afe-p-surface-uncertainty-2", "shape": "pair", "sideA": { "modality": "text", "value": "Clarifying question", "short": "Clarifying question" }, "sideB": { "modality": "text", "value": "A request for missing information that changes the task", "short": "A request for missing information that changes the task" }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "vocabulary" ], "uid": "h2wn4o1moyn00" }, { "id": "afe-d-surface-uncertainty", "shape": "definition", "term": { "modality": "text", "value": "Calibration", "short": "Calibration" }, "definition": { "modality": "text", "value": "How closely stated confidence corresponds to observed accuracy.", "short": "How closely stated confidence corresponds to observed accuracy." }, "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "definition" ], "uid": "lgel9x4v5wzr" }, { "id": "afe-q-surface-uncertainty-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should happen when a missing fact changes the answer?" }, "options": [ { "modality": "text", "value": "Hide the gap", "short": "Hide the gap" }, { "modality": "text", "value": "Invent a typical value", "short": "Invent a typical value" }, { "modality": "text", "value": "Ask a clarifying question", "short": "Ask to clarify" }, { "modality": "text", "value": "Claim full confidence", "short": "Claim full confidence" } ], "correctIndex": 2, "explanation": "Material missing information should be surfaced before proceeding.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "quiz" ], "uid": "wjaxq61hzfp02" }, { "id": "afe-q-surface-uncertainty-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "When is “80% confident” meaningful?" }, "options": [ { "modality": "text", "value": "Only in a long answer", "short": "Only in a long answer" }, { "modality": "text", "value": "After task-specific calibration", "short": "After calibrated tests" }, { "modality": "text", "value": "Whenever it sounds precise", "short": "Sounds precise" }, { "modality": "text", "value": "When no evidence exists", "short": "No evidence exists" } ], "correctIndex": 1, "explanation": "Confidence needs observed performance to become a reliable measure.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "quiz" ], "uid": "q4ehfxy3jbx" }, { "id": "afe-proc-surface-uncertainty", "shape": "procedure", "goal": "Create an assumption check", "steps": [ "List facts explicitly supplied by the user", "List inferences drawn from those facts", "Mark every unconfirmed premise as an assumption", "Ask about assumptions that would change the result", "Route unresolved uncertainty to verification or human review" ], "notes": "Do not turn a model's confidence wording into a probability without calibration evidence.", "source": { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, "tags": [ "roles-reasoning", "surface-uncertainty", "procedure" ], "uid": "goyw611sfgiyb" }, { "id": "afe-f-agent-loop-1", "shape": "fact", "title": "Agents Operate Across Steps", "body": "A goal-directed agent can select tools, inspect results, update state, and choose a next action instead of waiting for a new user message each time.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A goal-directed agent can select tools, inspect results, update state, and choose a next action instead of waiting for a new user message each time.", "illustration": { "kind": "photo", "imageSearchTerm": "project planning meeting sticky notes", "imagePrompt": "Editorial documentary photograph illustrating workflow loop planning tools. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A circular workflow of planning, action, checking, and stopping", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Refreshment_session_at_the_wikidata_training_in_lagos.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-loop-1.webp" }, "uid": "jq312uiw038e" }, { "id": "afe-f-agent-loop-2", "shape": "fact", "title": "Software Holds the Boundaries", "body": "The agent harness commonly supplies tools, permissions, memory, logs, and stop logic around the model.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "practical-ai" ], "illustration": { "imageSearchTerm": "fenced control panel boundary", "imagePrompt": "A sturdy fenced frame surrounding an open control panel, representing a software harness containing and bounding a system.", "alt": "Software Holds the Boundaries", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Mar_Lodge_Estate_-_Lui_Bridge_Plantation_(02NOV12)_(02).jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-loop-2.webp" }, "uid": "h7wh47glv9a9" }, { "id": "afe-f-agent-loop-3", "shape": "fact", "title": "Every Loop Needs an Exit", "body": "A success condition plus limits on attempts, time, and cost keeps a stalled workflow from running indefinitely.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "practical-ai" ], "illustration": { "imageSearchTerm": "running track finish gate", "imagePrompt": "A circular running track seen from above with one clearly marked finish gate breaking the loop.", "alt": "Every Loop Needs an Exit", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Bild_Carsten_Neder_-_Weltrekordversuch.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-loop-3.webp" }, "uid": "55oowv1kt4wxx" }, { "id": "afe-p-agent-loop-1", "shape": "pair", "sideA": { "modality": "text", "value": "Agent observation", "short": "Agent observation" }, "sideB": { "modality": "text", "value": "The result returned after an action", "short": "The result returned after an action" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "vocabulary" ], "uid": "zzty9onhghs0" }, { "id": "afe-p-agent-loop-2", "shape": "pair", "sideA": { "modality": "text", "value": "Stop condition", "short": "Stop condition" }, "sideB": { "modality": "text", "value": "A rule that ends the agent loop", "short": "A rule that ends the agent loop" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "vocabulary" ], "uid": "1vkwkew19ghtwo" }, { "id": "afe-d-agent-loop", "shape": "definition", "term": { "modality": "text", "value": "AI agent", "short": "AI agent" }, "definition": { "modality": "text", "value": "A goal-directed system that can choose and execute actions across multiple steps.", "short": "A goal-directed system that can choose and execute actions across multiple steps." }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "definition" ], "uid": "1pmthef187qcit" }, { "id": "afe-q-agent-loop-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What most distinguishes an agent?" }, "options": [ { "modality": "text", "value": "A humanlike avatar", "short": "A humanlike avatar" }, { "modality": "text", "value": "Multi-step action toward a goal", "short": "Multi-step goal action" }, { "modality": "text", "value": "A long context window", "short": "A long context window" }, { "modality": "text", "value": "A friendly chat tone", "short": "A friendly chat tone" } ], "correctIndex": 1, "explanation": "Agency comes from a loop that selects and executes actions.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "quiz" ], "uid": "2qfdv0yscopk" }, { "id": "afe-q-agent-loop-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What prevents an endless loop?" }, "options": [ { "modality": "text", "value": "A hidden greeting", "short": "A hidden greeting" }, { "modality": "text", "value": "An explicit stop condition", "short": "Explicit stop rule" }, { "modality": "text", "value": "A longer role prompt", "short": "A longer role prompt" }, { "modality": "text", "value": "More adjectives", "short": "More adjectives" } ], "correctIndex": 1, "explanation": "The workflow needs success and limit rules.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "quiz" ], "uid": "1gkc8jm6m5lvq" }, { "id": "afe-proc-agent-loop", "shape": "procedure", "goal": "Bound an agent loop", "steps": [ "State the goal and definition of done", "List allowed tools and actions", "Set attempt, time, and cost limits", "Check each observation against the goal", "Stop, escalate, or revise when a limit is reached" ], "notes": "A loop that cannot explain when it should stop is not ready to run unattended.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-loop", "procedure" ], "uid": "o5g6ax14mos0v" }, { "id": "afe-f-tools-permissions-1", "shape": "fact", "title": "Tools Create Real Effects", "body": "A model response becomes consequential when connected to search, files, databases, code, purchases, or communication channels.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A model response becomes consequential when connected to search, files, databases, code, purchases, or communication channels.", "illustration": { "kind": "photo", "imageSearchTerm": "keys permissions access control", "imagePrompt": "Editorial documentary photograph illustrating keys permissions access control. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A small set of labeled keys representing limited access", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Alcatraz,_Wikiexp_32.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tools-permissions-1.webp" }, "uid": "q1wel5smqga7" }, { "id": "afe-f-tools-permissions-2", "shape": "fact", "title": "State Records the Journey", "body": "Workflow state can track inputs, actions, observations, approvals, and remaining work across steps.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "practical-ai" ], "illustration": { "imageSearchTerm": "open logbook journal pen", "imagePrompt": "An open logbook with a pen resting across its pages, entries stacking chronologically down the page, no visible text.", "alt": "State Records the Journey", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Bay_side_of_hill,_path_down_to_water_side,_more_art_(7010503879).jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tools-permissions-2.webp" }, "uid": "hgr9e11kunjk3" }, { "id": "afe-f-tools-permissions-3", "shape": "fact", "title": "Least Privilege Limits Damage", "body": "Grant only the data and actions needed for the current task, for only as long as needed.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "practical-ai" ], "illustration": { "imageSearchTerm": "single key narrow door", "imagePrompt": "A single small key on a ring opening one narrow door, while a large ring of many other keys hangs locked away untouched.", "alt": "Least Privilege Limits Damage", "depictable": false, "credit": "Pexels · A vibrant image of a red locker door with a key in the lock, featuring bold primary colors.", "creditUrl": "https://www.pexels.com/photo/colorful-locker-with-key-in-bright-red-door-33009524/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tools-permissions-3.webp" }, "uid": "5j9y2m1duldou" }, { "id": "afe-p-tools-permissions-1", "shape": "pair", "sideA": { "modality": "text", "value": "Read permission", "short": "Read permission" }, "sideB": { "modality": "text", "value": "Ability to inspect without changing data", "short": "Ability to inspect without changing data" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "vocabulary" ], "uid": "y9bn4gjoprpc" }, { "id": "afe-p-tools-permissions-2", "shape": "pair", "sideA": { "modality": "text", "value": "Write permission", "short": "Write permission" }, "sideB": { "modality": "text", "value": "Ability to create or modify data", "short": "Ability to create or modify data" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "vocabulary" ], "uid": "1imgh6f1w0xtpp" }, { "id": "afe-d-tools-permissions", "shape": "definition", "term": { "modality": "text", "value": "Least privilege", "short": "Least privilege" }, "definition": { "modality": "text", "value": "Giving a user or system only the minimum access needed for its task.", "short": "Giving a user or system only the minimum access needed for its task." }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "definition" ], "uid": "46ixc117105ob" }, { "id": "afe-q-tools-permissions-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What access should a summarizer get?" }, "options": [ { "modality": "text", "value": "Read access to the chosen files", "short": "Read chosen files" }, { "modality": "text", "value": "All company passwords", "short": "All company passwords" }, { "modality": "text", "value": "Delete access to every folder", "short": "Delete every folder" }, { "modality": "text", "value": "Payment authority", "short": "Payment authority" } ], "correctIndex": 0, "explanation": "The task requires reading selected inputs, not broad authority.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "quiz" ], "uid": "ky0dauxfk03q" }, { "id": "afe-q-tools-permissions-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why separate draft from send?" }, "options": [ { "modality": "text", "value": "Drafting uses no language", "short": "Draft uses no language" }, { "modality": "text", "value": "Drafts cannot be reviewed", "short": "No draft review" }, { "modality": "text", "value": "Sending improves accuracy", "short": "Send boosts accuracy" }, { "modality": "text", "value": "Sending has an external effect", "short": "Sending affects others" } ], "correctIndex": 3, "explanation": "A send action affects another person and merits stronger control.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "quiz" ], "uid": "btt9o7xfkoz9" }, { "id": "afe-proc-tools-permissions", "shape": "procedure", "goal": "Set permissions for an agent", "steps": [ "List every required resource", "Separate read, create, update, send, and delete actions", "Remove capabilities the goal does not need", "Require approval for consequential operations", "Revoke temporary access when the task ends" ], "notes": "Permission boundaries should be enforced by the surrounding system, not only requested in prose.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "tools-permissions", "procedure" ], "uid": "lcmwp6b1tsn6" }, { "id": "afe-f-approval-checkpoints-1", "shape": "fact", "title": "Review Belongs Before Consequence", "body": "An approval is most valuable before an action becomes costly, external, legally significant, or difficult to reverse.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "An approval is most valuable before an action becomes costly, external, legally significant, or difficult to reverse.", "illustration": { "kind": "photo", "imageSearchTerm": "human approval checkpoint workflow", "imagePrompt": "Editorial documentary photograph illustrating human approval checkpoint workflow. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A person approving one stage before a workflow continues", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:01764_4_layers_Human_Tower_2011_Rio_de_Janeiro.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-approval-checkpoints-1.webp" }, "uid": "ng9iho1hf9wew" }, { "id": "afe-f-approval-checkpoints-2", "shape": "fact", "title": "Reviewers Need Evidence", "body": "A useful checkpoint shows what will happen, why, which inputs support it, and what uncertainty remains.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "practical-ai" ], "illustration": { "imageSearchTerm": "reviewer evidence folder desk", "imagePrompt": "A reviewer at a desk examining a folder of supporting evidence and printed charts before signing off on a decision.", "alt": "Reviewers Need Evidence", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-approval-checkpoints-2.webp" }, "uid": "14u2om91oh4i0n" }, { "id": "afe-f-approval-checkpoints-3", "shape": "fact", "title": "Automation Can Be Tiered", "body": "Low-risk reversible work may proceed automatically while sensitive, destructive, financial, or public actions require explicit approval.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "practical-ai" ], "illustration": { "imageSearchTerm": "traffic light tiered lanes", "imagePrompt": "A traffic light system with green, yellow, and red lanes leading to different gates, representing tiered levels of automated approval.", "alt": "Automation Can Be Tiered", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Dusky_-_Flickr_-_heiwa4126.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-approval-checkpoints-3.webp" }, "uid": "1ef5mr71dwv11x" }, { "id": "afe-p-approval-checkpoints-1", "shape": "pair", "sideA": { "modality": "text", "value": "Approval gate", "short": "Approval gate" }, "sideB": { "modality": "text", "value": "A required human decision before an action", "short": "A required human decision before an action" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "vocabulary" ], "uid": "7zj1dy1adbega" }, { "id": "afe-p-approval-checkpoints-2", "shape": "pair", "sideA": { "modality": "text", "value": "Reversible action", "short": "Reversible action" }, "sideB": { "modality": "text", "value": "A change that can be safely undone", "short": "A change that can be safely undone" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "vocabulary" ], "uid": "1arl1eq1e4cnum" }, { "id": "afe-d-approval-checkpoints", "shape": "definition", "term": { "modality": "text", "value": "Human-in-the-loop", "short": "Human-in-the-loop" }, "definition": { "modality": "text", "value": "A workflow in which a person reviews, directs, or approves selected system actions.", "short": "A workflow in which a person reviews" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "definition" ], "uid": "108efxu1cs6poe" }, { "id": "afe-q-approval-checkpoints-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which action most needs approval?" }, "options": [ { "modality": "text", "value": "Renaming a scratch heading", "short": "Rename scratch heading" }, { "modality": "text", "value": "Sending a payment", "short": "Sending a payment" }, { "modality": "text", "value": "Suggesting file tags", "short": "Suggesting file tags" }, { "modality": "text", "value": "Sorting local draft notes", "short": "Sort local draft notes" } ], "correctIndex": 1, "explanation": "A payment has financial consequence and may be hard to reverse.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "quiz" ], "uid": "11gwm0i1nbd15a" }, { "id": "afe-q-approval-checkpoints-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What makes review meaningful?" }, "options": [ { "modality": "text", "value": "No time to inspect", "short": "No time to inspect" }, { "modality": "text", "value": "Evidence and power to reject", "short": "Evidence + veto power" }, { "modality": "text", "value": "A preselected yes button", "short": "Preselected yes button" }, { "modality": "text", "value": "Hidden action details", "short": "Hidden action details" } ], "correctIndex": 1, "explanation": "The reviewer needs context, time, and genuine authority.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "quiz" ], "uid": "1x8gqz81ox6g20" }, { "id": "afe-cmp-approval-checkpoints", "shape": "comparison", "prompt": "Which checkpoint gives real control?", "correct": { "caption": "Show evidence and pause until a person approves or rejects.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-approval-checkpoints-show-evidence-pause.svg" }, "incorrect": { "caption": "Send first, then notify a person that review was available.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-approval-checkpoints-send-first-notify.svg" }, "explanation": "Approval must occur before the consequential action.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "approval-checkpoints", "comparison" ], "uid": "18ran5ihf9g2q" }, { "id": "afe-f-agent-failures-1", "shape": "fact", "title": "Errors Compound Across Steps", "body": "A mistaken plan or contaminated observation can influence every later decision in a multi-step workflow.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A mistaken plan or contaminated observation can influence every later decision in a multi-step workflow.", "illustration": { "kind": "photo", "imageSearchTerm": "workflow error recovery control room", "imagePrompt": "Editorial documentary photograph illustrating workflow error recovery control room. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A monitored workflow showing an error and recovery path", "credit": "Pexels · panumas nikhomkhai · Pexels License", "creditUrl": "https://www.pexels.com/photo/computer-server-in-data-center-room-17489158/", "subject": "A rack-mounted KVM console in a data centre, keyboard drawer pulled out, monitor showing a solid blue screen with no text, shot with a fisheye lens and tinted blue throughout", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-failures-1.webp" }, "uid": "1aozigkdsverc" }, { "id": "afe-f-agent-failures-2", "shape": "fact", "title": "External Content Can Be Hostile", "body": "A webpage, email, or document may contain instructions intended to redirect an agent. Treat retrieved content as data, not trusted control text.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "practical-ai" ], "illustration": { "imageSearchTerm": "padlock laptop incoming documents", "imagePrompt": "A closed padlock resting on a laptop keyboard beside a stack of incoming printed emails and papers, symbolizing content held at a distance before being trusted.", "alt": "External Content Can Be Hostile", "depictable": false, "credit": "Pexels · Flat lay of tax documents, smartphone calculator, and laptop for self-employment accounting.", "creditUrl": "https://www.pexels.com/photo/a-smartphone-with-a-calculator-app-over-a-tax-document-8962445/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-failures-2.webp" }, "uid": "1nd2uc3184p93h" }, { "id": "afe-f-agent-failures-3", "shape": "fact", "title": "Logs Enable Reconstruction", "body": "Action and approval logs help reviewers see what happened, diagnose failure, and improve the workflow.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "practical-ai" ], "illustration": { "imageSearchTerm": "computer terminal log screen", "imagePrompt": "A close-up of a computer monitor displaying rows of scrolling terminal log text at an angle, in a dim room, with no legible text visible.", "alt": "Logs Enable Reconstruction", "depictable": true, "credit": "Pexels · A classic MS-DOS terminal screen displayed on a laptop keyboard with vivid illumination.", "creditUrl": "https://www.pexels.com/photo/retro-ms-dos-terminal-with-keyboard-illumination-37657434/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-agent-failures-3.webp" }, "uid": "11noxmq1n2o9jq" }, { "id": "afe-p-agent-failures-1", "shape": "pair", "sideA": { "modality": "text", "value": "Agent drift", "short": "Agent drift" }, "sideB": { "modality": "text", "value": "Movement away from the user's intended goal", "short": "Movement away from the user's intended goal" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "vocabulary" ], "uid": "f6h6podyrwgc" }, { "id": "afe-p-agent-failures-2", "shape": "pair", "sideA": { "modality": "text", "value": "Escalation path", "short": "Escalation path" }, "sideB": { "modality": "text", "value": "A route to human help when the workflow cannot proceed safely", "short": "A route to human help when the workflow cannot proceed safely" }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "vocabulary" ], "uid": "a90dk91cakljv" }, { "id": "afe-d-agent-failures", "shape": "definition", "term": { "modality": "text", "value": "Prompt injection", "short": "Prompt injection" }, "definition": { "modality": "text", "value": "Untrusted content designed to override or redirect an AI system's instructions.", "short": "Untrusted content designed to override or redirect an AI system's instructions." }, "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "definition" ], "uid": "24yw8x1s1jai7" }, { "id": "afe-q-agent-failures-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should retrieved web text be treated?" }, "options": [ { "modality": "text", "value": "As permission to act", "short": "As permission to act" }, { "modality": "text", "value": "As untrusted data", "short": "As untrusted data" }, { "modality": "text", "value": "As highest-priority instructions", "short": "Top-priority orders" }, { "modality": "text", "value": "As verified truth", "short": "As verified truth" } ], "correctIndex": 1, "explanation": "External text may be inaccurate or adversarial.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "quiz" ], "uid": "1k3v42kmq2hp8" }, { "id": "afe-q-agent-failures-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why cap repeated tool calls?" }, "options": [ { "modality": "text", "value": "Guarantee perfect planning", "short": "Guarantee perfect plan" }, { "modality": "text", "value": "Hide errors from logs", "short": "Hide errors from logs" }, { "modality": "text", "value": "Increase every permission", "short": "Raise every permission" }, { "modality": "text", "value": "Stop wasteful failure loops", "short": "Stop failure loops" } ], "correctIndex": 3, "explanation": "Retry limits bound cost and force escalation.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "quiz" ], "uid": "dv34bk1nvk5o" }, { "id": "afe-proc-agent-failures", "shape": "procedure", "goal": "Test an agent before wider use", "steps": [ "Run representative normal tasks", "Simulate missing and malformed tool results", "Insert hostile instructions in retrieved content", "Verify logs, limits, and approval gates", "Expand authority only after failures are contained" ], "notes": "Test the control system around the model, not just the quality of its prose.", "source": { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, "tags": [ "agents", "agent-failures", "procedure" ], "uid": "1w1tex41vq4ob4" }, { "id": "afe-f-image-prompts-1", "shape": "fact", "title": "Start with the Visual Subject", "body": "State what must appear, what it is doing, and the relationship among important elements before adding style language.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "State what must appear, what it is doing, and the relationship among important elements before adding style language.", "illustration": { "kind": "photo", "imageSearchTerm": "photography moodboard composition lighting", "imagePrompt": "Editorial documentary photograph illustrating photography moodboard composition lighting. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A visual moodboard for subject, composition, and light", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Mohamed_2016_Tuulo_ku_dhow_Boorama..jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-image-prompts-1.webp" }, "uid": "50q6lpkhoqwf" }, { "id": "afe-f-image-prompts-2", "shape": "fact", "title": "Composition Directs Attention", "body": "Viewpoint, framing, scale, subject placement, depth, and negative space determine how the eye moves through an image.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "practical-ai" ], "illustration": { "imageSearchTerm": "Composition Directs Attention", "imagePrompt": "Composition Directs Attention. Viewpoint, framing, scale, subject placement, depth, and negative space determine how the eye moves through an image.", "alt": "Composition Directs Attention", "credit": "Pexels · Ambient Walking", "creditUrl": "https://www.pexels.com/photo/top-view-of-a-road-sign-5391955/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-image-prompts-2.webp" }, "uid": "1gfcydy1m7d6a6" }, { "id": "afe-f-image-prompts-3", "shape": "fact", "title": "Iteration Beats Keyword Piles", "body": "Inspect a result and revise a specific visual dimension. Many fashionable adjectives can compete without clarifying the composition.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "practical-ai" ], "illustration": { "imageSearchTerm": "artist easel canvas adjustment", "imagePrompt": "An artist adjusting the lighting on a single canvas on an easel, with a cluttered mood board of too many competing color swatches pinned up behind.", "alt": "Iteration Beats Keyword Piles", "depictable": false, "credit": "Pexels · Artist painting on an easel in a black and white setting, with a paintbrush and canvas.", "creditUrl": "https://www.pexels.com/photo/an-artist-painting-on-a-canvas-21077233/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-image-prompts-3.webp" }, "uid": "1mdlnwweoquv4" }, { "id": "afe-p-image-prompts-1", "shape": "pair", "sideA": { "modality": "text", "value": "Image subject", "short": "Image subject" }, "sideB": { "modality": "text", "value": "The main person, object, or scene", "short": "The main person" }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "vocabulary" ], "uid": "1tf0m9v5bful5" }, { "id": "afe-p-image-prompts-2", "shape": "pair", "sideA": { "modality": "text", "value": "Negative prompt", "short": "Negative prompt" }, "sideB": { "modality": "text", "value": "A description of elements to avoid", "short": "A description of elements to avoid" }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "vocabulary" ], "uid": "1yfg3f1g9mmm5" }, { "id": "afe-d-image-prompts", "shape": "definition", "term": { "modality": "text", "value": "Creative brief", "short": "Creative brief" }, "definition": { "modality": "text", "value": "A concise statement of a visual project's goal, subject, audience, and constraints.", "short": "A concise statement of a visual project's goal" }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "definition" ], "uid": "1syt4xn1ae1n6h" }, { "id": "afe-q-image-prompts-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should an image prompt name first?" }, "options": [ { "modality": "text", "value": "An invisible watermark", "short": "An invisible watermark" }, { "modality": "text", "value": "A random camera brand", "short": "A random camera brand" }, { "modality": "text", "value": "The essential subject", "short": "The essential subject" }, { "modality": "text", "value": "Twenty style buzzwords", "short": "Twenty style buzzwords" } ], "correctIndex": 2, "explanation": "The required visual content should lead the brief.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "quiz" ], "uid": "1t3zoxr14c7uqx" }, { "id": "afe-q-image-prompts-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "A portrait feels cramped. Best revision?" }, "options": [ { "modality": "text", "value": "Request factual citations", "short": "Request citations" }, { "modality": "text", "value": "Increase text length", "short": "Increase text length" }, { "modality": "text", "value": "Add more art styles", "short": "Add more art styles" }, { "modality": "text", "value": "Ask for more negative space", "short": "Ask for negative space" } ], "correctIndex": 3, "explanation": "Negative space directly addresses the framing problem.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "quiz" ], "uid": "ydsazk1p3mo08" }, { "id": "afe-proc-image-prompts", "shape": "procedure", "goal": "Revise a generated image", "steps": [ "Identify the one visual problem that matters most", "Name the desired change in composition, light, color, or subject", "Preserve elements that already work", "Generate a small set of variants", "Check details, rights, and intended use before publishing" ], "notes": "Keep the revision specific enough that you can tell whether it succeeded.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "image-prompts", "procedure" ], "uid": "icsuvq19oh9oq" }, { "id": "afe-f-aspect-ratios-1", "shape": "fact", "title": "Aspect Ratio Shapes Composition", "body": "The width-to-height relationship changes where subjects, motion, text space, and visual balance can fit.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "aspect-ratios", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "The width-to-height relationship changes where subjects, motion, text space, and visual balance can fit.", "illustration": { "kind": "photo", "imageSearchTerm": "gallery wall picture frames interior", "imagePrompt": "Editorial documentary photograph illustrating photo aspect ratio frames. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Several image frames showing square, portrait, and landscape canvases", "credit": "Pexels · A cozy interior with a gallery wall of framed pictures and soft ambient lighting.", "creditUrl": "https://www.pexels.com/photo/wooden-picture-frames-hanging-on-the-wall-7877826/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-aspect-ratios-1.webp" }, "uid": "7055a91rm10cr" }, { "id": "afe-f-aspect-ratios-2", "shape": "fact", "title": "Design for the Destination", "body": "Choose the canvas for the feed, slide, poster, thumbnail, or banner where the asset will actually appear.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "aspect-ratios", "practical-ai" ], "illustration": { "imageSearchTerm": "picture frames various sizes", "imagePrompt": "Several empty picture frames and canvases of different aspect ratios, tall, wide, and square, leaning together against a wall.", "alt": "Design for the Destination", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Picture_in_Picture_mode.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-aspect-ratios-2.webp" }, "uid": "7maik2jw1fpm" }, { "id": "afe-f-aspect-ratios-3", "shape": "fact", "title": "Cropping Can Remove Meaning", "body": "Converting one ratio to another may cut faces, hands, products, or context. Preview each required crop rather than assuming it is safe.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "aspect-ratios", "practical-ai" ], "illustration": { "imageSearchTerm": "photo crop frame overlay", "imagePrompt": "A photograph on a light table with a crop-frame overlay tool positioned over it, showing the edges would cut off part of the subject.", "alt": "Cropping Can Remove Meaning", "depictable": true }, "uid": "1ljmx2vbwn6wh" }, { "id": "afe-n-aspect-ratios-1", "shape": "numeric", "prompt": { "modality": "text", "value": "Square canvas" }, "value": 1, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "w2hcbp1kizgan" }, { "id": "afe-n-aspect-ratios-2", "shape": "numeric", "prompt": { "modality": "text", "value": "Portrait 3:4 canvas" }, "value": 0.75, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "fcmvjz1mofj0h" }, { "id": "afe-n-aspect-ratios-3", "shape": "numeric", "prompt": { "modality": "text", "value": "Landscape 4:3 canvas" }, "value": 1.333, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "1w0czcdhtb8zr" }, { "id": "afe-n-aspect-ratios-4", "shape": "numeric", "prompt": { "modality": "text", "value": "Vertical 9:16 canvas" }, "value": 0.5625, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "1dicpy91ef3x5n" }, { "id": "afe-n-aspect-ratios-5", "shape": "numeric", "prompt": { "modality": "text", "value": "Widescreen 16:9 canvas" }, "value": 1.778, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "n644uvlrrtdh" }, { "id": "afe-n-aspect-ratios-6", "shape": "numeric", "prompt": { "modality": "text", "value": "Classic photo 3:2 canvas" }, "value": 1.5, "unit": "width ÷ height", "tolerance": 0.01, "source": { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, "tags": [ "multimodal", "aspect-ratios", "aspect-ratio", "numeric" ], "uid": "hdpmkq1npxjwy" }, { "id": "afe-q-aspect-ratios-special", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which canvas is tallest relative to width?" }, "options": [ { "modality": "text", "value": "9:16 vertical", "short": "9:16 vertical" }, { "modality": "text", "value": "16:9 widescreen", "short": "16:9 widescreen" }, { "modality": "text", "value": "4:3 landscape", "short": "4:3 landscape" }, { "modality": "text", "value": "1:1 square", "short": "1:1 square" } ], "correctIndex": 0, "explanation": "Nine-by-sixteen has the smallest width-to-height value in this set.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "aspect-ratios", "quiz" ], "uid": "2rwfg3q8xvkd" }, { "id": "afe-f-video-planning-1", "shape": "fact", "title": "Video Prompts Describe Change", "body": "Action, camera movement, duration, pacing, and transitions matter because the output unfolds over time.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Action, camera movement, duration, pacing, and transitions matter because the output unfolds over time.", "illustration": { "kind": "photo", "imageSearchTerm": "film storyboard sketches desk", "imagePrompt": "Editorial documentary photograph illustrating video storyboard shot sequence. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Storyboard frames arranged into a short scene", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Captura_de_pantalla_2025-11-04_a_las_23.42.20.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-video-planning-1.webp" }, "uid": "1hcrc3m1hy28dq" }, { "id": "afe-f-video-planning-2", "shape": "fact", "title": "Storyboards Externalize the Plan", "body": "A sequence of frames or shot descriptions lets you inspect narrative beats and composition before generating footage.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "practical-ai" ], "illustration": { "imageSearchTerm": "storyboard sketch panels wall", "imagePrompt": "A hand-drawn storyboard with a grid of sequential sketch panels pinned to a wall, showing a visual sequence with no legible text.", "alt": "Storyboards Externalize the Plan", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Almost_Vuture_Cover_Page.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-video-planning-2.webp" }, "uid": "xuzzl2lddi96" }, { "id": "afe-f-video-planning-3", "shape": "fact", "title": "Continuity Needs Deliberate Control", "body": "Characters, props, lighting, geography, and direction can drift between clips unless they are specified and checked.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "practical-ai" ], "illustration": { "imageSearchTerm": "film set continuity photos", "imagePrompt": "A film set continuity binder open on a table, showing several reference photos of the same prop and costume pinned together for comparison, no legible text.", "alt": "Continuity Needs Deliberate Control", "depictable": true, "credit": "Pexels · Black and white photo of hands holding a clapperboard on a film set, capturing cinematic action.", "creditUrl": "https://www.pexels.com/photo/a-person-holding-a-clipper-board-21716757/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-video-planning-3.webp" }, "uid": "ouf9butb48my" }, { "id": "afe-p-video-planning-1", "shape": "pair", "sideA": { "modality": "text", "value": "Shot list", "short": "Shot list" }, "sideB": { "modality": "text", "value": "An ordered plan of camera setups", "short": "An ordered plan of camera setups" }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "vocabulary" ], "uid": "rn6cb6mi0m9m" }, { "id": "afe-p-video-planning-2", "shape": "pair", "sideA": { "modality": "text", "value": "Continuity", "short": "Continuity" }, "sideB": { "modality": "text", "value": "Consistency of visual details across shots", "short": "Consistency of visual details across shots" }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "vocabulary" ], "uid": "nuzm8c1hqn96g" }, { "id": "afe-d-video-planning", "shape": "definition", "term": { "modality": "text", "value": "Storyboard", "short": "Storyboard" }, "definition": { "modality": "text", "value": "A sequence of images or panels that plans how a video will unfold.", "short": "A sequence of images or panels that plans how a video will unfold." }, "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "definition" ], "uid": "h9gciamttsn2" }, { "id": "afe-q-video-planning-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What does a video prompt add beyond a still?" }, "options": [ { "modality": "text", "value": "Change over time", "short": "Change over time" }, { "modality": "text", "value": "Guaranteed continuity", "short": "Guaranteed continuity" }, { "modality": "text", "value": "A single aspect ratio only", "short": "One aspect ratio only" }, { "modality": "text", "value": "Automatic consent", "short": "Automatic consent" } ], "correctIndex": 0, "explanation": "Motion and duration are defining dimensions of video.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "quiz" ], "uid": "l2z6b4cpz0w" }, { "id": "afe-q-video-planning-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Best way to control a long sequence?" }, "options": [ { "modality": "text", "value": "Ignore screen direction", "short": "Ignore shot direction" }, { "modality": "text", "value": "Use one vague prompt", "short": "Use one vague prompt" }, { "modality": "text", "value": "Remove all references", "short": "Remove all references" }, { "modality": "text", "value": "Generate and edit short shots", "short": "Edit short shots" } ], "correctIndex": 3, "explanation": "Short, planned clips make selection and continuity checks easier.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "quiz" ], "uid": "bzb1ye13rhysi" }, { "id": "afe-proc-video-planning", "shape": "procedure", "goal": "Plan a generated video sequence", "steps": [ "Write the audience and single message", "Break the message into visual beats", "Specify each shot's subject, action, camera, and duration", "Lock continuity details and reference frames", "Generate short takes, review, and edit" ], "notes": "Sound, captions, rights, and accessibility belong in the production plan too.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "video-planning", "procedure" ], "uid": "zk4rlp8jidn" }, { "id": "afe-f-audio-workflows-1", "shape": "fact", "title": "Audio Tasks Are Not Interchangeable", "body": "Transcription, speaker labeling, translation, voice synthesis, music, and effects use different inputs and require different reviews.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Transcription, speaker labeling, translation, voice synthesis, music, and effects use different inputs and require different reviews.", "illustration": { "kind": "photo", "imageSearchTerm": "audio waveform studio microphone", "imagePrompt": "Editorial documentary photograph illustrating audio waveform studio microphone. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A studio microphone beside an audio waveform display", "credit": "Pexels · Close-up of a hand working in an audio recording studio with microphone and monitor.", "creditUrl": "https://www.pexels.com/photo/black-flat-screen-computer-monitor-on-the-table-6892721/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-audio-workflows-1.webp" }, "uid": "oy0b531m67dh5" }, { "id": "afe-f-audio-workflows-2", "shape": "fact", "title": "Transcripts Need Human Checks", "body": "Names, numbers, technical terms, accents, noise, and overlapping speakers are common places for speech recognition to fail.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "practical-ai" ], "illustration": { "imageSearchTerm": "headphones transcript audio review", "imagePrompt": "A person wearing headphones at a desk, reviewing a printed transcript next to a computer screen showing an audio waveform, no legible text.", "alt": "Transcripts Need Human Checks", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:A-Line_freebees_from_opening_weekend.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-audio-workflows-2.webp" }, "uid": "qgqu1nntrirh" }, { "id": "afe-f-audio-workflows-3", "shape": "fact", "title": "A Voice Can Be an Identity Signal", "body": "Synthetic speech can imitate a person, enabling accessibility and creativity as well as impersonation and fraud.", "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "practical-ai" ], "illustration": { "imageSearchTerm": "microphone sound wave recording", "imagePrompt": "A studio microphone with a glowing sound waveform visualization behind it in a dim recording booth.", "alt": "A Voice Can Be an Identity Signal", "depictable": true, "credit": "Pexels · Thomas Lineweaver · Pexels License", "creditUrl": "https://www.pexels.com/photo/greyscale-photography-of-condenser-microphone-765139/", "subject": "Black-and-white photo of a broadcast microphone in a shock mount on a boom arm, with a monitor behind showing multi-track audio waveforms; illegible sticky notes on the monitor bezel", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-audio-workflows-3.webp" }, "uid": "13sfsbeedbifm" }, { "id": "afe-k-audio-workflows-1", "shape": "concept", "conceptKind": "thing", "name": "Transcription", "clues": [ "Turns recorded speech into written text.", "Names and numbers need special review.", "Its output is a transcript." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "1ukkpfz17a1m99" }, { "id": "afe-k-audio-workflows-2", "shape": "concept", "conceptKind": "thing", "name": "Diarization", "clues": [ "Works with recordings containing multiple speakers.", "Answers the question of who spoke when.", "It labels speaker turns." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "yt9n091yo47ab" }, { "id": "afe-k-audio-workflows-3", "shape": "concept", "conceptKind": "thing", "name": "Speech translation", "clues": [ "Changes spoken language for another audience.", "It combines recognition and translation.", "Its goal is speech in another language." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "1oiuoo26ny9pq" }, { "id": "afe-k-audio-workflows-4", "shape": "concept", "conceptKind": "thing", "name": "Voice synthesis", "clues": [ "Produces spoken audio from text or controls.", "It may use a generic or imitated voice.", "Consent matters when identity is copied." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "umz85ngrqxvd" }, { "id": "afe-k-audio-workflows-5", "shape": "concept", "conceptKind": "thing", "name": "Music generation", "clues": [ "Creates rhythmic and tonal audio.", "Prompts may name instrumentation, tempo, and mood.", "Its output is music." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "ehb2j2194npfi" }, { "id": "afe-k-audio-workflows-6", "shape": "concept", "conceptKind": "thing", "name": "Sound-effect generation", "clues": [ "Creates short nonmusical audio events.", "Prompts name source, environment, and duration.", "Footsteps or a door close are examples." ], "source": { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, "tags": [ "multimodal", "audio-workflows", "key-concept" ], "uid": "1234izgzom8e4" }, { "id": "afe-q-audio-workflows-special", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which task labels who spoke when?" }, "options": [ { "modality": "text", "value": "Speech translation", "short": "Speech translation" }, { "modality": "text", "value": "Transcription", "short": "Transcription" }, { "modality": "text", "value": "Voice synthesis", "short": "Voice synthesis" }, { "modality": "text", "value": "Diarization", "short": "Diarization" } ], "correctIndex": 3, "explanation": "Diarization identifies and separates speaker turns.", "source": { "label": "Google Cloud — Speaker diarization", "url": "https://docs.cloud.google.com/speech-to-text/docs/multiple-voices" }, "tags": [ "multimodal", "audio-workflows", "quiz" ], "uid": "1c9mzmp7fgt8j" }, { "id": "afe-f-draft-edit-1", "shape": "fact", "title": "Drafting and Editing Are Different Jobs", "body": "Drafting creates a first version from a brief. Editing diagnoses or changes existing language according to stated goals.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Drafting creates a first version from a brief. Editing diagnoses or changes existing language according to stated goals.", "illustration": { "kind": "photo", "imageSearchTerm": "red pen editing printed paper", "imagePrompt": "Editorial documentary photograph illustrating writer editing printed draft. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A writer marking revisions on a printed draft", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:20140701_Bucure%C5%9Fti_01.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-draft-edit-1.webp" }, "uid": "1frv2nu1xbjmxi" }, { "id": "afe-f-draft-edit-2", "shape": "fact", "title": "Preservation Rules Prevent Drift", "body": "Identify facts, quotations, commitments, voice, and other elements that must survive a rewrite.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "practical-ai" ], "illustration": { "imageSearchTerm": "red pen editing preserved sentence", "imagePrompt": "A red pen circling one sentence left untouched on a page while the surrounding sentences show heavy editing marks, no legible text.", "alt": "Preservation Rules Prevent Drift", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:%22_12_-_ITALY_-_Joe_Colombo_-_Tube_Chair_-_modularity_and_flexibility_in_modular_structure_Italy_1969_red.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-draft-edit-2.webp" }, "uid": "lujk8k17o5mds" }, { "id": "afe-f-draft-edit-3", "shape": "fact", "title": "A Change Log Makes Revision Visible", "body": "For consequential documents, a short list of material edits helps the author inspect what changed and why.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "practical-ai" ], "illustration": { "imageSearchTerm": "tracked changes document revisions", "imagePrompt": "A document on a screen showing tracked-changes markup with colored insertion and deletion marks along the margin, no legible text.", "alt": "A Change Log Makes Revision Visible", "depictable": true }, "uid": "1mnieura9ww3l" }, { "id": "afe-p-draft-edit-1", "shape": "pair", "sideA": { "modality": "text", "value": "Developmental edit", "short": "Developmental edit" }, "sideB": { "modality": "text", "value": "Changes structure, argument, or coverage", "short": "Changes structure" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "vocabulary" ], "uid": "l50da5el9myz" }, { "id": "afe-p-draft-edit-2", "shape": "pair", "sideA": { "modality": "text", "value": "Copy edit", "short": "Copy edit" }, "sideB": { "modality": "text", "value": "Improves clarity, consistency, and correctness", "short": "Improves clarity" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "vocabulary" ], "uid": "pr1nir1blllf5" }, { "id": "afe-d-draft-edit", "shape": "definition", "term": { "modality": "text", "value": "Revision brief", "short": "Revision brief" }, "definition": { "modality": "text", "value": "Instructions describing what to change, what to preserve, and how the result will be judged.", "short": "Instructions describing what to change" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "definition" ], "uid": "1ft6tlw1oc9228" }, { "id": "afe-q-draft-edit-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should a rewrite prompt protect?" }, "options": [ { "modality": "text", "value": "Meaning that must not change", "short": "Preserve core meaning" }, { "modality": "text", "value": "Every original typo", "short": "Every original typo" }, { "modality": "text", "value": "Only the word count", "short": "Only the word count" }, { "modality": "text", "value": "A random persona", "short": "A random persona" } ], "correctIndex": 0, "explanation": "Preservation rules keep fluent revision from changing substance.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "quiz" ], "uid": "1froaytypw1rz" }, { "id": "afe-q-draft-edit-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Best first step for sensitive prose?" }, "options": [ { "modality": "text", "value": "Hide all changes", "short": "Hide all changes" }, { "modality": "text", "value": "Ask for diagnosis before rewriting", "short": "Diagnose before edit" }, { "modality": "text", "value": "Remove the original", "short": "Remove the original" }, { "modality": "text", "value": "Replace it without review", "short": "Replace without review" } ], "correctIndex": 1, "explanation": "A diagnostic pass lets the author choose which edits to accept.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "quiz" ], "uid": "8ups7vbcaj6t" }, { "id": "afe-p-draft-edit-3", "shape": "pair", "sideA": { "modality": "text", "value": "Change log", "short": "Change log" }, "sideB": { "modality": "text", "value": "A concise record of material revisions", "short": "A concise record of material revisions" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "draft-edit", "vocabulary" ], "uid": "1293jfv62xfa9" }, { "id": "afe-f-summarize-1", "shape": "fact", "title": "A Summary Has a Purpose", "body": "Audience and decision need determine which details are central, which are supporting, and which can be omitted.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Audience and decision need determine which details are central, which are supporting, and which can be omitted.", "illustration": { "kind": "photo", "imageSearchTerm": "stack reports summary sticky note", "imagePrompt": "Editorial documentary photograph illustrating long report summary notes. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A long report reduced to a concise set of notes", "credit": "Pexels · Stack of colorful sticky notes with a 'To Do' note on top, isolated on white.", "creditUrl": "https://www.pexels.com/photo/a-to-do-note-on-a-stack-of-sticky-note-pads-6991379/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-summarize-1.webp" }, "uid": "tpxh0ymhrkvm" }, { "id": "afe-f-summarize-2", "shape": "fact", "title": "Long Context Is Not Perfect Recall", "body": "Research finds that model performance can vary with where relevant information appears, including degradation for material in the middle.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "practical-ai" ], "illustration": { "imageSearchTerm": "long scroll faded middle", "imagePrompt": "A long unrolled scroll of paper with the middle section faded and blurred while both ends remain sharp and clear.", "alt": "Long Context Is Not Perfect Recall", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-summarize-2.webp" }, "uid": "1fo20di195panq" }, { "id": "afe-f-summarize-3", "shape": "fact", "title": "Traceability Improves Review", "body": "Section, page, timestamp, or passage links let a reader compare summary claims with the original material.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "practical-ai" ], "illustration": { "imageSearchTerm": "sticky tabs source document", "imagePrompt": "A summary document with colored sticky tabs linking specific lines to matching highlighted passages in a thick source book beside it, no legible text.", "alt": "Traceability Improves Review", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Sticky.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-summarize-3.webp" }, "uid": "18qrszn1wyj19x" }, { "id": "afe-p-summarize-1", "shape": "pair", "sideA": { "modality": "text", "value": "Executive summary", "short": "Executive summary" }, "sideB": { "modality": "text", "value": "Decision-focused overview for a busy reader", "short": "Decision-focused overview for a busy reader" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "vocabulary" ], "uid": "5tc88m167hs7e" }, { "id": "afe-p-summarize-2", "shape": "pair", "sideA": { "modality": "text", "value": "Extractive summary", "short": "Extractive summary" }, "sideB": { "modality": "text", "value": "A selection of passages taken from the source", "short": "A selection of passages taken from the source" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "vocabulary" ], "uid": "wrxegu1umuvyu" }, { "id": "afe-d-summarize", "shape": "definition", "term": { "modality": "text", "value": "Abstractive summary", "short": "Abstractive summary" }, "definition": { "modality": "text", "value": "A condensed restatement written in new language rather than copied passages.", "short": "A condensed restatement written in new language rather than copied passages." }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "definition" ], "uid": "1f1u67f1ih9ofl" }, { "id": "afe-q-summarize-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should a summary prompt specify?" }, "options": [ { "modality": "text", "value": "A fictional credential", "short": "A fictional credential" }, { "modality": "text", "value": "Audience and coverage", "short": "Audience and coverage" }, { "modality": "text", "value": "No source boundaries", "short": "No source boundaries" }, { "modality": "text", "value": "Only “make shorter”", "short": "Only “make shorter”" } ], "correctIndex": 1, "explanation": "Selection depends on who needs what from the material.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "quiz" ], "uid": "wxyz9qhjqt2e" }, { "id": "afe-q-summarize-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How can a long-document summary be checked?" }, "options": [ { "modality": "text", "value": "Hide omitted material", "short": "Hide omitted material" }, { "modality": "text", "value": "Link claims to source locations", "short": "Link claims to sources" }, { "modality": "text", "value": "Remove section labels", "short": "Remove section labels" }, { "modality": "text", "value": "Trust confident wording", "short": "Trust confident prose" } ], "correctIndex": 1, "explanation": "Traceable references enable comparison with the original.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "quiz" ], "uid": "gf606ei32mpq" }, { "id": "afe-cmp-summarize", "shape": "comparison", "prompt": "Which summary request is more reliable?", "correct": { "caption": "Summarize these sections for a decision; cite page numbers.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-summarize-summarize-these-sections.svg" }, "incorrect": { "caption": "Read everything and tell me the important parts from memory.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-summarize-read-everything-tell.svg" }, "explanation": "Purpose and traceability make selection and review explicit.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "summarize", "comparison" ], "uid": "1494okf181nt6p" }, { "id": "afe-f-transform-1", "shape": "fact", "title": "Transformation Changes Representation", "body": "A format or tone change should preserve the underlying facts and intended meaning unless the brief explicitly requests substantive revision.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A format or tone change should preserve the underlying facts and intended meaning unless the brief explicitly requests substantive revision.", "illustration": { "kind": "photo", "imageSearchTerm": "documents multiple formats desk", "imagePrompt": "Editorial documentary photograph illustrating content formats email table slides. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "One message represented as an email, table, and slide outline", "credit": "Pexels · Kampus Production · Pexels License", "creditUrl": "https://www.pexels.com/photo/documents-in-a-binder-on-a-work-desk-8353764/", "subject": "A tidy white office desk with a dark monitor and keyboard; an open black portfolio holds a page of hand-drawn diagrams and sketches beside a typed table-style page, with two more loose printed forms on the desk; no legible text", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-transform-1.webp" }, "uid": "1a673od1i2pfdv" }, { "id": "afe-f-transform-2", "shape": "fact", "title": "Audience Defines Plain Language", "body": "Useful simplification depends on what the reader already knows and what action they need to take.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "practical-ai" ], "illustration": { "imageSearchTerm": "two readers different ages", "imagePrompt": "Two readers of different ages studying the same open picture book side by side, each reacting to it differently.", "alt": "Audience Defines Plain Language", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-transform-2.webp" }, "uid": "1qh4ijx1o8912v" }, { "id": "afe-f-transform-3", "shape": "fact", "title": "Fluency Can Hide Semantic Drift", "body": "A smoother rewrite may drop a caveat, change a quantity, or turn a possibility into a promise.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "practical-ai" ], "illustration": { "imageSearchTerm": "polished stone rough stone", "imagePrompt": "A smooth polished stone resting beside a rougher, slightly larger original stone, suggesting something changed in the polishing.", "alt": "Fluency Can Hide Semantic Drift", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-transform-3.webp" }, "uid": "1yav72f1fov5kh" }, { "id": "afe-p-transform-1", "shape": "pair", "sideA": { "modality": "text", "value": "Semantic invariant", "short": "Semantic invariant" }, "sideB": { "modality": "text", "value": "Meaning that must remain unchanged", "short": "Meaning that must remain unchanged" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "vocabulary" ], "uid": "aj6mf31h0eixx" }, { "id": "afe-p-transform-2", "shape": "pair", "sideA": { "modality": "text", "value": "Format transformation", "short": "Format transformation" }, "sideB": { "modality": "text", "value": "Recasting content into a new structure", "short": "Recasting content into a new structure" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "vocabulary" ], "uid": "1czw2xw2zfvn8" }, { "id": "afe-d-transform", "shape": "definition", "term": { "modality": "text", "value": "Tone", "short": "Tone" }, "definition": { "modality": "text", "value": "The attitude and relationship conveyed by choices of wording, rhythm, and emphasis.", "short": "The attitude and relationship conveyed by choices of wording" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "definition" ], "uid": "c8ervv9sk7uh" }, { "id": "afe-q-transform-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What must a format conversion preserve?" }, "options": [ { "modality": "text", "value": "Material facts and qualifications", "short": "Facts & qualifications" }, { "modality": "text", "value": "A random metaphor", "short": "A random metaphor" }, { "modality": "text", "value": "Every line break", "short": "Every line break" }, { "modality": "text", "value": "The model's first guess", "short": "Model's first guess" } ], "correctIndex": 0, "explanation": "Representation can change while substantive meaning remains fixed.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "quiz" ], "uid": "195x6whcyyp7v" }, { "id": "afe-q-transform-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should “friendly” be clarified?" }, "options": [ { "modality": "text", "value": "Repeat it ten times", "short": "Repeat it ten times" }, { "modality": "text", "value": "Name audience and relationship", "short": "Name audience/role" }, { "modality": "text", "value": "Ask for total certainty", "short": "Ask for certainty" }, { "modality": "text", "value": "Remove the purpose", "short": "Remove the purpose" } ], "correctIndex": 1, "explanation": "Tone becomes concrete when tied to a social situation.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "quiz" ], "uid": "o5e2r91kjigvf" }, { "id": "afe-p-transform-3", "shape": "pair", "sideA": { "modality": "text", "value": "Plain-language rewrite", "short": "Plain-language rewrite" }, "sideB": { "modality": "text", "value": "A version adapted for readers without specialist knowledge", "short": "A version adapted for readers without specialist knowledge" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "transform", "vocabulary" ], "uid": "4qaucce1r4j4" }, { "id": "afe-f-document-qa-1", "shape": "fact", "title": "Bound the Evidence Set", "body": "A document Q&A task should identify which files count as evidence and whether outside sources are allowed.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A document Q&A task should identify which files count as evidence and whether outside sources are allowed.", "illustration": { "kind": "photo", "imageSearchTerm": "person reading report desk highlighter", "imagePrompt": "Editorial documentary photograph illustrating document search highlighted passages. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "Documents with relevant passages highlighted for review", "credit": "Pexels · ANTONI SHKRABA production · Pexels License", "creditUrl": "https://www.pexels.com/photo/close-up-of-person-sitting-on-table-taking-notes-on-paper-8374275/", "subject": "Close-up of a hand in a tan jacket highlighting lines on a printed document with an orange highlighter, several lines already highlighted, more sheets beneath, laptop and a glass of water alongside", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-document-qa-1.webp" }, "uid": "sn7zd0p3qxz4" }, { "id": "afe-f-document-qa-2", "shape": "fact", "title": "Not Found Is Not False", "body": "Failure to locate a claim in a packet shows only that it was not found there; it does not establish that the claim is untrue everywhere.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "practical-ai" ], "illustration": { "imageSearchTerm": "filing cabinet unopened drawer", "imagePrompt": "A person searching through an open filing cabinet with one drawer left unopened, representing an unexplored possibility rather than a proven absence.", "alt": "Not Found Is Not False", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Wild_trim_and_nat_comm_2.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-document-qa-2.webp" }, "uid": "ph96zw1hvjdyw" }, { "id": "afe-f-document-qa-3", "shape": "fact", "title": "Schemas Make Extraction Testable", "body": "Named fields, types, and missing-value rules turn loose reading into output that can be sampled and compared with originals.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "practical-ai" ], "illustration": { "imageSearchTerm": "labeled form grid boxes", "imagePrompt": "A grid of small labeled boxes on a form being filled in one at a time from a document beside it, no legible text.", "alt": "Schemas Make Extraction Testable", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-document-qa-3.webp" }, "uid": "1mtj8xjyemh7" }, { "id": "afe-p-document-qa-1", "shape": "pair", "sideA": { "modality": "text", "value": "Evidence packet", "short": "Evidence packet" }, "sideB": { "modality": "text", "value": "The defined documents allowed to support an answer", "short": "The defined documents allowed to support an answer" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "vocabulary" ], "uid": "119n2fv11z0881" }, { "id": "afe-p-document-qa-2", "shape": "pair", "sideA": { "modality": "text", "value": "Missing value rule", "short": "Missing value rule" }, "sideB": { "modality": "text", "value": "How absent or unknown fields should be represented", "short": "How absent or unknown fields should be represented" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "vocabulary" ], "uid": "1nlcp6o1yh7kgk" }, { "id": "afe-d-document-qa", "shape": "definition", "term": { "modality": "text", "value": "Document grounding", "short": "Document grounding" }, "definition": { "modality": "text", "value": "Constraining an answer to information found in a supplied set of documents.", "short": "Constraining an answer to information found in a supplied set of documents." }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "definition" ], "uid": "3wf8vl1we2rs7" }, { "id": "afe-q-document-qa-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should grounded Q&A do when evidence is absent?" }, "options": [ { "modality": "text", "value": "Say it was not found", "short": "Say it was not found" }, { "modality": "text", "value": "Cite an unrelated page", "short": "Cite an unrelated page" }, { "modality": "text", "value": "Treat absence as false", "short": "Treat absence as false" }, { "modality": "text", "value": "Invent a likely answer", "short": "Invent a likely answer" } ], "correctIndex": 0, "explanation": "The packet may be incomplete, so absence must be stated narrowly.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "quiz" ], "uid": "1k2pr9k1ff0ios" }, { "id": "afe-q-document-qa-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What improves extraction review?" }, "options": [ { "modality": "text", "value": "Hidden source files", "short": "Hidden source files" }, { "modality": "text", "value": "A defined field schema", "short": "A defined field schema" }, { "modality": "text", "value": "A vague request", "short": "A vague request" }, { "modality": "text", "value": "No missing-value rule", "short": "No missing-value rule" } ], "correctIndex": 1, "explanation": "A schema provides explicit fields and expected representations.", "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "quiz" ], "uid": "1pnkwlp8n4bjj" }, { "id": "afe-p-document-qa-3", "shape": "pair", "sideA": { "modality": "text", "value": "Source citation", "short": "Source citation" }, "sideB": { "modality": "text", "value": "A pointer from an answer to supporting evidence", "short": "A pointer from an answer to supporting evidence" }, "source": { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, "tags": [ "writing-documents", "document-qa", "vocabulary" ], "uid": "fd8x8z1312wpl" }, { "id": "afe-f-research-plan-1", "shape": "fact", "title": "Research Starts with Claim Types", "body": "A historical date, medical recommendation, product price, and personal preference require different sources and levels of verification.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A historical date, medical recommendation, product price, and personal preference require different sources and levels of verification.", "illustration": { "kind": "photo", "imageSearchTerm": "library books research notes desk", "imagePrompt": "Editorial documentary photograph illustrating research plan library sources. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A research plan beside books and source notes", "credit": "Pexels · A young man in a library studying with books and taking notes under a desk lamp.", "creditUrl": "https://www.pexels.com/photo/a-man-studying-in-a-library-6550126/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-research-plan-1.webp" }, "uid": "7owivd1yqurbn" }, { "id": "afe-f-research-plan-2", "shape": "fact", "title": "Search Terms Are Hypotheses", "body": "Suggested queries help explore a topic, but wording can bias what appears. Try synonyms, counterarguments, and source-specific searches.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "practical-ai" ], "illustration": { "imageSearchTerm": "magnifying glass research page", "imagePrompt": "A magnifying glass hovering over a page of text with several different colored highlighter marks representing different search angles, no legible text.", "alt": "Search Terms Are Hypotheses", "depictable": false, "credit": "Pexels · Magnifying glass on educational papers with books and grass background, outdoors.", "creditUrl": "https://www.pexels.com/photo/close-up-shot-of-a-magnifying-glass-on-top-of-books-12719255/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-research-plan-2.webp" }, "uid": "b86rqp1yzin8z" }, { "id": "afe-f-research-plan-3", "shape": "fact", "title": "A Citation Must Support the Claim", "body": "A real link is not sufficient if its content does not entail, qualify, or directly document the sentence it is attached to.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "practical-ai" ], "illustration": { "imageSearchTerm": "evidence board connecting string", "imagePrompt": "A red string connecting a highlighted sentence in one document to a specific paragraph in a separate source document on a table, evidence-board style.", "alt": "A Citation Must Support the Claim", "depictable": false, "credit": "Pexels · A detailed crime investigation board filled with photos, maps, and red strings connecting clues and evidence.", "creditUrl": "https://www.pexels.com/photo/investigation-board-with-photos-maps-and-cutouts-showing-connections-8369512/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-research-plan-3.webp" }, "uid": "k4fpsm14fi9ua" }, { "id": "afe-p-research-plan-1", "shape": "pair", "sideA": { "modality": "text", "value": "Primary source", "short": "Primary source" }, "sideB": { "modality": "text", "value": "Original evidence or an authoritative first-party record", "short": "Original evidence or an authoritative first-party record" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "vocabulary" ], "uid": "10spum5n8r37n" }, { "id": "afe-p-research-plan-2", "shape": "pair", "sideA": { "modality": "text", "value": "Secondary source", "short": "Secondary source" }, "sideB": { "modality": "text", "value": "Analysis or synthesis based on other sources", "short": "Analysis or synthesis based on other sources" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "vocabulary" ], "uid": "1t78d37zwagg9" }, { "id": "afe-d-research-plan", "shape": "definition", "term": { "modality": "text", "value": "Evidence table", "short": "Evidence table" }, "definition": { "modality": "text", "value": "A structured record linking claims to sources, passages, dates, and confidence notes.", "short": "A structured record linking claims to sources" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "definition" ], "uid": "l37dgv1mp0pwd" }, { "id": "afe-q-research-plan-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should happen to an AI citation?" }, "options": [ { "modality": "text", "value": "Open and verify it", "short": "Open and verify it" }, { "modality": "text", "value": "Assume the quote exists", "short": "Assume quote exists" }, { "modality": "text", "value": "Count the URL as proof", "short": "Count the URL as proof" }, { "modality": "text", "value": "Trust the title alone", "short": "Trust the title alone" } ], "correctIndex": 0, "explanation": "The source and its support for the claim both need checking.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "quiz" ], "uid": "1y8m8331ptvl6l" }, { "id": "afe-q-research-plan-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why search counterarguments?" }, "options": [ { "modality": "text", "value": "Reduce one-sided evidence", "short": "Reduce one-sided proof" }, { "modality": "text", "value": "Avoid primary sources", "short": "Avoid primary sources" }, { "modality": "text", "value": "Guarantee consensus", "short": "Guarantee consensus" }, { "modality": "text", "value": "Make every answer longer", "short": "Make answers longer" } ], "correctIndex": 0, "explanation": "Deliberately testing alternatives helps reveal missing evidence.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "quiz" ], "uid": "l7mn4dxiec5z" }, { "id": "afe-p-research-plan-3", "shape": "pair", "sideA": { "modality": "text", "value": "Claim-source alignment", "short": "Claim-source alignment" }, "sideB": { "modality": "text", "value": "The source directly supports the attached statement", "short": "The source directly supports the attached statement" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "research-plan", "vocabulary" ], "uid": "o4yrg2lk4uyu" }, { "id": "afe-f-tutor-1", "shape": "fact", "title": "Practice Beats Passive Answering", "body": "Questions, retrieval, explanation, and correction require the learner to produce knowledge instead of only reading a generated solution.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Questions, retrieval, explanation, and correction require the learner to produce knowledge instead of only reading a generated solution.", "illustration": { "kind": "photo", "imageSearchTerm": "student practice flashcards feedback", "imagePrompt": "Editorial documentary photograph illustrating student practice flashcards feedback. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A learner practicing with questions and feedback", "credit": "Pexels · Andy Barbour · Pexels License", "creditUrl": "https://www.pexels.com/photo/person-checking-test-papers-6684372/", "subject": "Top-down close-up of two hands over a bubble-sheet titled 'Answer Sheet for Practice Test', one hand holding a red marker that has written a circled 'B' at the top; laptop edge and pencil nearby", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tutor-1.webp" }, "uid": "192r9gf14pvj91" }, { "id": "afe-f-tutor-2", "shape": "fact", "title": "Hints Preserve Productive Struggle", "body": "A staged hint can unblock progress without immediately replacing the learner's reasoning.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "practical-ai" ], "illustration": { "imageSearchTerm": "student hint card worksheet", "imagePrompt": "A student working through a handwritten problem on paper, with a small hint card partially visible nearby offering just one clue, no legible text.", "alt": "Hints Preserve Productive Struggle", "depictable": true, "credit": "Pexels · George Pak · Pexels License", "creditUrl": "https://www.pexels.com/photo/person-writing-on-white-paper-7972487/", "subject": "Seen over the shoulder of a long-haired student outdoors on paving, a pen in hand working through an algebra worksheet headed 'Solve the following equations', Sections A-E, with handwritten workings on the page", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tutor-2.webp" }, "uid": "1u2xhjx1bd71n7" }, { "id": "afe-f-tutor-3", "shape": "fact", "title": "Tutoring Output Needs a Reference", "body": "A model can give confident but incorrect feedback, so learning sessions should connect to a trusted text, teacher, or answer key.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "practical-ai" ], "illustration": { "imageSearchTerm": "teacher answer key checking", "imagePrompt": "A teacher cross-referencing a student's worksheet against a printed answer key at a desk, no legible text.", "alt": "Tutoring Output Needs a Reference", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:English_teacher.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-tutor-3.webp" }, "uid": "180yi47upnhhp" }, { "id": "afe-p-tutor-1", "shape": "pair", "sideA": { "modality": "text", "value": "Retrieval practice", "short": "Retrieval practice" }, "sideB": { "modality": "text", "value": "Recalling knowledge without looking at the answer", "short": "Recalling knowledge without looking at the answer" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "vocabulary" ], "uid": "171xaoc1uw5vac" }, { "id": "afe-p-tutor-2", "shape": "pair", "sideA": { "modality": "text", "value": "Scaffolded hint", "short": "Scaffolded hint" }, "sideB": { "modality": "text", "value": "A hint that preserves the learner's next step", "short": "A hint that preserves the learner's next step" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "vocabulary" ], "uid": "1yg5qaw114wmsw" }, { "id": "afe-d-tutor", "shape": "definition", "term": { "modality": "text", "value": "Formative feedback", "short": "Formative feedback" }, "definition": { "modality": "text", "value": "Information intended to improve learning during practice rather than only score the final result.", "short": "Information intended to improve learning during practice rather than only score the final result." }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "definition" ], "uid": "ltj97edbe1z6" }, { "id": "afe-q-tutor-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which tutor prompt keeps the learner active?" }, "options": [ { "modality": "text", "value": "Give every answer immediately", "short": "Give answers at once" }, { "modality": "text", "value": "Wait for my attempt first", "short": "Wait for my attempt" }, { "modality": "text", "value": "Remove all questions", "short": "Remove all questions" }, { "modality": "text", "value": "Skip feedback", "short": "Skip feedback" } ], "correctIndex": 1, "explanation": "An attempt creates evidence of what the learner understands.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "quiz" ], "uid": "pwge511fz46b" }, { "id": "afe-q-tutor-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should AI feedback be treated?" }, "options": [ { "modality": "text", "value": "Replace the course source", "short": "Replace course source" }, { "modality": "text", "value": "Assume confidence means truth", "short": "Confidence isn't truth" }, { "modality": "text", "value": "Check it against trusted material", "short": "Check trusted material" }, { "modality": "text", "value": "Hide disagreements", "short": "Hide disagreements" } ], "correctIndex": 2, "explanation": "Generated feedback can be useful and mistaken.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "quiz" ], "uid": "nseqc11a1nq2r" }, { "id": "afe-p-tutor-3", "shape": "pair", "sideA": { "modality": "text", "value": "Socratic question", "short": "Socratic question" }, "sideB": { "modality": "text", "value": "A prompt that advances thinking without supplying the conclusion", "short": "A prompt that advances thinking without supplying the conclusion" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "tutor", "vocabulary" ], "uid": "y6io5b1q0uugt" }, { "id": "afe-f-decision-support-1", "shape": "fact", "title": "Criteria Encode Values", "body": "A decision table reflects what the chooser counts, how it is weighted, and which tradeoffs are acceptable.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A decision table reflects what the chooser counts, how it is weighted, and which tradeoffs are acceptable.", "illustration": { "kind": "photo", "imageSearchTerm": "decision matrix comparison table", "imagePrompt": "A hand filling in a decision matrix grid on paper, comparing several options against multiple weighted criteria, no legible text.", "alt": "A decision matrix with options, criteria, and tradeoffs", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Coutts-gonet2.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decision-support-1.webp" }, "uid": "zlxpwpri9len" }, { "id": "afe-f-decision-support-2", "shape": "fact", "title": "Scores Need Evidence", "body": "A neat matrix can manufacture precision if ratings are guesses. Link each material score to data, experience, or an explicit assumption.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "practical-ai" ], "illustration": { "imageSearchTerm": "scoring sheet supporting evidence", "imagePrompt": "A scoring sheet with numeric ratings on a table, each row clipped to a small stack of supporting evidence documents, no legible text.", "alt": "Scores Need Evidence", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:IMS_scoring_pylon_20080510.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decision-support-2.webp" }, "uid": "kwgbw01vsgwbo" }, { "id": "afe-f-decision-support-3", "shape": "fact", "title": "Sensitivity Reveals Fragility", "body": "If a small change in one weight reverses the recommendation, the decision depends heavily on that judgment.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "practical-ai" ], "illustration": { "imageSearchTerm": "balance scale tipping weight", "imagePrompt": "A finely balanced scale tipping dramatically to one side after a single small weight is added to it.", "alt": "Sensitivity Reveals Fragility", "depictable": false, "credit": "Pexels · Black and white image of a vintage balance scale in an İstanbul museum display.", "creditUrl": "https://www.pexels.com/photo/balance-scale-on-display-in-the-museum-15253932/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-decision-support-3.webp" }, "uid": "bnu5vw1ycoj3k" }, { "id": "afe-p-decision-support-1", "shape": "pair", "sideA": { "modality": "text", "value": "Must-have criterion", "short": "Must-have criterion" }, "sideB": { "modality": "text", "value": "A requirement an option cannot trade away", "short": "A requirement an option cannot trade away" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "vocabulary" ], "uid": "1gm1ocdpk6x2n" }, { "id": "afe-p-decision-support-2", "shape": "pair", "sideA": { "modality": "text", "value": "Sensitivity analysis", "short": "Sensitivity analysis" }, "sideB": { "modality": "text", "value": "Testing how conclusions change when assumptions change", "short": "Testing how conclusions change when assumptions change" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "vocabulary" ], "uid": "1qxwl2yhbzqii" }, { "id": "afe-d-decision-support", "shape": "definition", "term": { "modality": "text", "value": "Decision matrix", "short": "Decision matrix" }, "definition": { "modality": "text", "value": "A table that compares options across explicitly chosen criteria.", "short": "A table that compares options across explicitly chosen criteria." }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "definition" ], "uid": "xtqkxbtbeokl" }, { "id": "afe-q-decision-support-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What makes a decision matrix deceptive?" }, "options": [ { "modality": "text", "value": "Named criteria", "short": "Named criteria" }, { "modality": "text", "value": "Visible tradeoffs", "short": "Visible tradeoffs" }, { "modality": "text", "value": "Source links", "short": "Source links" }, { "modality": "text", "value": "Unsupported precise scores", "short": "Unsupported scores" } ], "correctIndex": 3, "explanation": "Numbers can create false precision when they lack evidence.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "quiz" ], "uid": "1ywfhx81jjal3w" }, { "id": "afe-q-decision-support-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What does sensitivity analysis test?" }, "options": [ { "modality": "text", "value": "Whether prose is friendly", "short": "Friendly prose" }, { "modality": "text", "value": "Whether a model is conscious", "short": "Model consciousness" }, { "modality": "text", "value": "Whether files are private", "short": "File privacy" }, { "modality": "text", "value": "Whether the ranking is fragile", "short": "Ranking fragility" } ], "correctIndex": 3, "explanation": "It varies assumptions or weights and observes the result.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "quiz" ], "uid": "197i8161xbacca" }, { "id": "afe-p-decision-support-3", "shape": "pair", "sideA": { "modality": "text", "value": "Decision owner", "short": "Decision owner" }, "sideB": { "modality": "text", "value": "The person accountable for the final choice", "short": "The person accountable for the final choice" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "decision-support", "vocabulary" ], "uid": "17nlul47mz21w" }, { "id": "afe-f-thinking-moves-1", "shape": "fact", "title": "Thinking Has Modes", "body": "Generating possibilities, reducing information, comparing alternatives, challenging claims, reframing problems, and sequencing action are different tasks.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Generating possibilities, reducing information, comparing alternatives, challenging claims, reframing problems, and sequencing action are different tasks.", "illustration": { "kind": "photo", "imageSearchTerm": "assorted tools table variety", "imagePrompt": "Six different physical tools, a magnifying glass, a funnel, a set of scales, a prism, a picture frame, and an arrow, arranged neatly on a table, each suggesting a different way of thinking.", "alt": "Cards representing brainstorming, critique, comparison, and planning", "depictable": false, "credit": "AI-generated (gpt-image-1.5)", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-thinking-moves-1.webp" }, "uid": "andng716pdpwx" }, { "id": "afe-f-thinking-moves-2", "shape": "fact", "title": "Separate Expansion from Selection", "body": "Brainstorming and judging at the same time can narrow the search too early. Generate first, then apply criteria.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "practical-ai" ], "illustration": { "imageSearchTerm": "sticky notes wall brainstorm", "imagePrompt": "A wide spray of scattered sticky notes covering a wall, with only a small cluster circled in a later, separate pass, no legible text.", "alt": "Separate Expansion from Selection", "depictable": false, "credit": "Pexels · Yellow sticky notes with motivational business ideas on a white wall.", "creditUrl": "https://www.pexels.com/photo/sticky-notes-on-the-white-wall-6914470/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-thinking-moves-2.webp" }, "uid": "ot7pnwg3osd8" }, { "id": "afe-f-thinking-moves-3", "shape": "fact", "title": "Lists Are Not Decisions", "body": "AI can widen and organize a choice set, but a person remains responsible for values, evidence, and final commitment.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "practical-ai" ], "illustration": { "imageSearchTerm": "clipboard checklist pen pause", "imagePrompt": "A long checklist on a clipboard held out, with a pen hovering just above it as a hand pauses before making the final mark.", "alt": "Lists Are Not Decisions", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Iclippy.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-thinking-moves-3.webp" }, "uid": "1t4qgblhf8gxn" }, { "id": "afe-k-thinking-moves-1", "shape": "concept", "conceptKind": "thing", "name": "Brainstorming", "clues": [ "Expands the option space before judgment.", "Quantity and variety matter at first.", "It generates possibilities." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "5l4mfe1ymref2" }, { "id": "afe-k-thinking-moves-2", "shape": "concept", "conceptKind": "thing", "name": "Summarizing", "clues": [ "Compresses a larger body of material.", "It selects what matters for an audience.", "It creates a shorter representation." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "gehty41k7qzf8" }, { "id": "afe-k-thinking-moves-3", "shape": "concept", "conceptKind": "thing", "name": "Comparing", "clues": [ "Places alternatives against shared criteria.", "It makes similarities and differences visible.", "It examines options side by side." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "1xn0epwbikrko" }, { "id": "afe-k-thinking-moves-4", "shape": "concept", "conceptKind": "thing", "name": "Critiquing", "clues": [ "Searches for weaknesses, omissions, and counterarguments.", "It pressure-tests an idea.", "Its job is constructive challenge." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "x8rybazjmzoy" }, { "id": "afe-k-thinking-moves-5", "shape": "concept", "conceptKind": "thing", "name": "Reframing", "clues": [ "Changes the question or point of view.", "It can reveal a hidden assumption.", "It offers a different way to define the problem." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "1h56jxwh6b1i4" }, { "id": "afe-k-thinking-moves-6", "shape": "concept", "conceptKind": "thing", "name": "Sequencing", "clues": [ "Turns a goal into an ordered path.", "Dependencies and checkpoints matter.", "It arranges actions over time." ], "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "key-concept" ], "uid": "144f270120ch44" }, { "id": "afe-q-thinking-moves-special", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which move deliberately seeks weaknesses?" }, "options": [ { "modality": "text", "value": "Sequencing", "short": "Sequencing" }, { "modality": "text", "value": "Brainstorming", "short": "Brainstorming" }, { "modality": "text", "value": "Summarizing", "short": "Summarizing" }, { "modality": "text", "value": "Critiquing", "short": "Critiquing" } ], "correctIndex": 3, "explanation": "Critiquing pressure-tests claims and looks for omissions.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "research-learning", "thinking-moves", "quiz" ], "uid": "1ynhkff109brxx" }, { "id": "afe-f-prepare-data-1", "shape": "fact", "title": "Rows Need a Unit of Observation", "body": "State what one row represents. Confusing a transaction with a customer or a reading with a day changes every calculation.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "State what one row represents. Confusing a transaction with a customer or a reading with a day changes every calculation.", "illustration": { "kind": "photo", "imageSearchTerm": "spreadsheet rows data table", "imagePrompt": "A close-up of a spreadsheet on a laptop screen showing many rows and columns of numerical data, no legible text.", "alt": "A clean data table beside a field schema", "depictable": true, "credit": "Unsplash · KOBU Agency · Unsplash License", "creditUrl": "https://unsplash.com/photos/text-on-white-background-Yyd_WQWyIyI", "subject": "on-screen table of coronavirus cases by US state with headers Total Cases / New Cases / Total Deaths / New Deaths / Active Cases and rows for New York, New Jersey, Michigan, California, Louisiana", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-prepare-data-1.webp" }, "uid": "8blypa4u6pke" }, { "id": "afe-f-prepare-data-2", "shape": "fact", "title": "A Data Dictionary Prevents Guessing", "body": "Field definitions, units, allowed values, and missing markers give both the model and reviewer a shared interpretation.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "practical-ai" ], "illustration": { "imageSearchTerm": "data dictionary reference sheet", "imagePrompt": "A printed reference sheet listing structured field definitions beside an open spreadsheet on a laptop, no legible text.", "alt": "A Data Dictionary Prevents Guessing", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Co-generated_data.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-prepare-data-2.webp" }, "uid": "tsid6p1bgw5df" }, { "id": "afe-f-prepare-data-3", "shape": "fact", "title": "Profile Before Analysis", "body": "Counts, types, ranges, missingness, duplicates, and invalid values reveal whether a dataset is ready for a question.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "practical-ai" ], "illustration": { "imageSearchTerm": "data dashboard summary charts", "imagePrompt": "A computer screen showing a data profiling dashboard with several summary charts, histograms, and counters, no legible text.", "alt": "Profile Before Analysis", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Citibike_daily_ridership.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-prepare-data-3.webp" }, "uid": "pyyhkd1e5kwjr" }, { "id": "afe-p-prepare-data-1", "shape": "pair", "sideA": { "modality": "text", "value": "Data dictionary", "short": "Data dictionary" }, "sideB": { "modality": "text", "value": "Definitions and rules for dataset fields", "short": "Definitions and rules for dataset fields" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "vocabulary" ], "uid": "gaf7pvjtirrt" }, { "id": "afe-p-prepare-data-2", "shape": "pair", "sideA": { "modality": "text", "value": "Unit of observation", "short": "Unit of observation" }, "sideB": { "modality": "text", "value": "The entity or event represented by one row", "short": "The entity or event represented by one row" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "vocabulary" ], "uid": "4mzl5v1ydxh91" }, { "id": "afe-d-prepare-data", "shape": "definition", "term": { "modality": "text", "value": "Schema", "short": "Schema" }, "definition": { "modality": "text", "value": "A structured description of fields, types, relationships, and constraints.", "short": "A structured description of fields" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "definition" ], "uid": "md17or1b1hy39" }, { "id": "afe-q-prepare-data-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should be stated about each row?" }, "options": [ { "modality": "text", "value": "The model's persona", "short": "The model's persona" }, { "modality": "text", "value": "A hidden conclusion", "short": "A hidden conclusion" }, { "modality": "text", "value": "What it represents", "short": "What it represents" }, { "modality": "text", "value": "A preferred chart color", "short": "Preferred chart color" } ], "correctIndex": 2, "explanation": "The unit of observation determines valid aggregation.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "quiz" ], "uid": "1s81fvom5csfc" }, { "id": "afe-q-prepare-data-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What comes before a trend claim?" }, "options": [ { "modality": "text", "value": "Invent missing values", "short": "Invent missing values" }, { "modality": "text", "value": "Choose the conclusion", "short": "Choose the conclusion" }, { "modality": "text", "value": "Profile and clean the data", "short": "Profile & clean data" }, { "modality": "text", "value": "Delete the original", "short": "Delete the original" } ], "correctIndex": 2, "explanation": "Basic quality checks protect the analysis from structural errors.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "quiz" ], "uid": "veltdk13v86j8" }, { "id": "afe-p-prepare-data-3", "shape": "pair", "sideA": { "modality": "text", "value": "Missingness rate", "short": "Missingness rate" }, "sideB": { "modality": "text", "value": "The share of values absent in a field", "short": "The share of values absent in a field" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "prepare-data", "vocabulary" ], "uid": "3gcc3yjcvk7m" }, { "id": "afe-f-formulas-charts-1", "shape": "fact", "title": "Formulas Depend on Context", "body": "Cell locations, table names, locale, date formats, blank handling, and software dialect can change the correct expression.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Cell locations, table names, locale, date formats, blank handling, and software dialect can change the correct expression.", "illustration": { "kind": "photo", "imageSearchTerm": "spreadsheet formula cell reference", "imagePrompt": "A close-up of a spreadsheet cell with a formula selected, its referenced cells outlined in matching colors, no legible text.", "alt": "A spreadsheet formula being checked against a chart", "depictable": true, "credit": "Pexels · Kindel Media · Pexels License", "creditUrl": "https://www.pexels.com/photo/monthly-budget-planning-7054399/", "subject": "Flat lay on white: a printed Monthly Budget report with a donut chart and a small bar chart, a white calculator reading 0, and four pastel pens, all in the lower right with large empty space at left.", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-formulas-charts-1.webp" }, "uid": "xrs8f0n11320" }, { "id": "afe-f-formulas-charts-2", "shape": "fact", "title": "Tiny Test Cases Catch Big Errors", "body": "A few rows with known outputs can reveal off-by-one ranges, wrong absolute references, and incorrect treatment of blanks.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "practical-ai" ], "illustration": { "imageSearchTerm": "small spreadsheet test rows", "imagePrompt": "A small handful of spreadsheet rows highlighted separately from a much larger dataset, being checked closely on a laptop screen, no legible text.", "alt": "Tiny Test Cases Catch Big Errors", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Image_taken_from_page_124_of_%27La_Terra,_trattato_popolare_di_geografia_universale_per_G._Marinelli_ed_altri_scienziati_italiani,_etc._(With_illustrations_and_maps.)%27_(11159467355).jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-formulas-charts-2.webp" }, "uid": "13a0u5m1sd707m" }, { "id": "afe-f-formulas-charts-3", "shape": "fact", "title": "Charts Encode Analytical Choices", "body": "Aggregation, denominator, interval, axis scale, and grouping determine what a visual comparison actually means.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "practical-ai" ], "illustration": { "imageSearchTerm": "bar chart line chart screen", "imagePrompt": "A computer screen displaying several different chart types, a bar chart, a line chart, and a pie chart, built from the same dataset, no legible text.", "alt": "Charts Encode Analytical Choices", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Cocktail_Blue_Lagoon_at_the_pitch_bar.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-formulas-charts-3.webp" }, "uid": "1lmnuhk1eq2huw" }, { "id": "afe-p-formulas-charts-1", "shape": "pair", "sideA": { "modality": "text", "value": "Absolute reference", "short": "Absolute reference" }, "sideB": { "modality": "text", "value": "A spreadsheet reference fixed when copied", "short": "A spreadsheet reference fixed when copied" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "vocabulary" ], "uid": "1m13ich1dyfjaz" }, { "id": "afe-p-formulas-charts-2", "shape": "pair", "sideA": { "modality": "text", "value": "Aggregation", "short": "Aggregation" }, "sideB": { "modality": "text", "value": "Combining values with a rule such as sum or average", "short": "Combining values with a rule such as sum or average" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "vocabulary" ], "uid": "veugcqavqc22" }, { "id": "afe-d-formulas-charts", "shape": "definition", "term": { "modality": "text", "value": "Data validation", "short": "Data validation" }, "definition": { "modality": "text", "value": "Rules that restrict or check values entered into a field or cell.", "short": "Rules that restrict or check values entered into a field or cell." }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "definition" ], "uid": "1lwww54q0u09s" }, { "id": "afe-q-formulas-charts-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should an AI formula be checked?" }, "options": [ { "modality": "text", "value": "Hide the inputs", "short": "Hide the inputs" }, { "modality": "text", "value": "Trust syntax alone", "short": "Trust syntax alone" }, { "modality": "text", "value": "Delete the source data", "short": "Delete the source data" }, { "modality": "text", "value": "Test known small cases", "short": "Test known small cases" } ], "correctIndex": 3, "explanation": "Hand-computed examples show whether the formula means what you intend.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "quiz" ], "uid": "ke890n9jqy25" }, { "id": "afe-q-formulas-charts-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should a chart request identify?" }, "options": [ { "modality": "text", "value": "Question and aggregation", "short": "Question + aggregation" }, { "modality": "text", "value": "Only a color palette", "short": "Only a color palette" }, { "modality": "text", "value": "A fictional expert", "short": "A fictional expert" }, { "modality": "text", "value": "No units", "short": "No units" } ], "correctIndex": 0, "explanation": "The analytical question determines the appropriate grouping and visual form.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "quiz" ], "uid": "xeyvbn1gdw4zh" }, { "id": "afe-p-formulas-charts-3", "shape": "pair", "sideA": { "modality": "text", "value": "Chart annotation", "short": "Chart annotation" }, "sideB": { "modality": "text", "value": "Text that explains a notable value or event", "short": "Text that explains a notable value or event" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "formulas-charts", "vocabulary" ], "uid": "gdhoaex0b3lm" }, { "id": "afe-f-coding-cycle-1", "shape": "fact", "title": "Repository Context Matters", "body": "Correct code depends on versions, APIs, conventions, neighboring modules, and tests that are not visible in an isolated snippet.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Correct code depends on versions, APIs, conventions, neighboring modules, and tests that are not visible in an isolated snippet.", "illustration": { "kind": "photo", "imageSearchTerm": "programmer coding laptop screen", "imagePrompt": "A close-up of a programmer's hands typing on a laptop keyboard, with colorful syntax-highlighted code visible on the screen, no legible text.", "alt": "A developer reviewing a small code change and its tests", "depictable": true, "credit": "Pexels · Person coding at a desk with laptop and external monitor showing programming code.", "creditUrl": "https://www.pexels.com/photo/a-person-doing-computer-programming-7988086/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-coding-cycle-1.webp" }, "uid": "1brfsf414oa6qo" }, { "id": "afe-f-coding-cycle-2", "shape": "fact", "title": "Small Diffs Are Easier to Trust", "body": "A narrowly scoped patch is simpler to review, test, revert, and compare with the requested behavior.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "practical-ai" ], "illustration": { "imageSearchTerm": "code diff red green lines", "imagePrompt": "A code editor screen showing a small diff view with a few highlighted green added lines and red removed lines, no legible text.", "alt": "Small Diffs Are Easier to Trust", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:GSoC_Book_Jotter.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-coding-cycle-2.webp" }, "uid": "1bz0so71rwf7px" }, { "id": "afe-f-coding-cycle-3", "shape": "fact", "title": "Execution Provides Evidence", "body": "Formatting, types, tests, builds, and targeted runtime checks reveal problems that plausible-looking code can hide.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "practical-ai" ], "illustration": { "imageSearchTerm": "terminal running tests screen", "imagePrompt": "A dark terminal window on a monitor showing a scrolling test run with green checkmarks and a few red failure indicators, no legible text.", "alt": "Execution Provides Evidence", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:2013-06-08_Projekt_Hei%C3%9Flufftballon_DSCF7636.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-coding-cycle-3.webp" }, "uid": "oj4v6kxnxrrg" }, { "id": "afe-p-coding-cycle-1", "shape": "pair", "sideA": { "modality": "text", "value": "Patch", "short": "Patch" }, "sideB": { "modality": "text", "value": "A focused set of code changes", "short": "A focused set of code changes" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "vocabulary" ], "uid": "1erzxoa1lameq6" }, { "id": "afe-p-coding-cycle-2", "shape": "pair", "sideA": { "modality": "text", "value": "Regression test", "short": "Regression test" }, "sideB": { "modality": "text", "value": "A test that prevents a fixed bug from returning", "short": "A test that prevents a fixed bug from returning" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "vocabulary" ], "uid": "2wny5yjsooeq" }, { "id": "afe-d-coding-cycle", "shape": "definition", "term": { "modality": "text", "value": "Code review", "short": "Code review" }, "definition": { "modality": "text", "value": "Systematic inspection of a change for correctness, clarity, security, and fit.", "short": "Systematic inspection of a change for correctness" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "definition" ], "uid": "16cr2wmpb0uaa" }, { "id": "afe-q-coding-cycle-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should precede a broad code edit?" }, "options": [ { "modality": "text", "value": "Assume the API version", "short": "Assume the API version" }, { "modality": "text", "value": "Remove the tests", "short": "Remove the tests" }, { "modality": "text", "value": "Rewrite every file", "short": "Rewrite every file" }, { "modality": "text", "value": "Inspect context and plan", "short": "Inspect context + plan" } ], "correctIndex": 3, "explanation": "The change must fit the real repository and its constraints.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "quiz" ], "uid": "1raqkil9hfbdr" }, { "id": "afe-q-coding-cycle-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why prefer a small patch?" }, "options": [ { "modality": "text", "value": "It guarantees no bugs", "short": "It guarantees no bugs" }, { "modality": "text", "value": "It ignores conventions", "short": "It ignores conventions" }, { "modality": "text", "value": "It is easier to review and revert", "short": "Review + easy rollback" }, { "modality": "text", "value": "It needs no tests", "short": "It needs no tests" } ], "correctIndex": 2, "explanation": "Limited scope improves inspection and recovery.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "quiz" ], "uid": "15ropdgjgehes" }, { "id": "afe-p-coding-cycle-3", "shape": "pair", "sideA": { "modality": "text", "value": "Static check", "short": "Static check" }, "sideB": { "modality": "text", "value": "Analysis performed without running the program", "short": "Analysis performed without running the program" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "coding-cycle", "vocabulary" ], "uid": "17egln11on2q7f" }, { "id": "afe-f-test-secure-1", "shape": "fact", "title": "Generated Code Is Untrusted Input", "body": "Fluent code should enter the same review, testing, dependency, and security process as code from any other untrusted source.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Fluent code should enter the same review, testing, dependency, and security process as code from any other untrusted source.", "illustration": { "kind": "photo", "imageSearchTerm": "software security testing terminal", "imagePrompt": "Editorial documentary photograph illustrating software security testing terminal. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A software test suite and security review in progress", "credit": "Unsplash · Jefferson Santos · Unsplash License", "creditUrl": "https://unsplash.com/photos/person-using-laptop-computers-9SoCnyQmkzI", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-test-secure-1.webp" }, "uid": "gwwuatfi8ezn" }, { "id": "afe-f-test-secure-2", "shape": "fact", "title": "Happy Paths Are Not Enough", "body": "Boundary, malformed, failure, concurrent, and unauthorized cases reveal behavior that a normal example never exercises.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "practical-ai" ], "illustration": { "imageSearchTerm": "smooth path rough paths", "imagePrompt": "A single smooth, well-lit path through a garden beside several rougher, obstacle-filled paths branching off into shadow.", "alt": "Happy Paths Are Not Enough", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Jay_Smooth.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-test-secure-2.webp" }, "uid": "15lkkpl7smtqf" }, { "id": "afe-f-test-secure-3", "shape": "fact", "title": "Security Needs Independent Controls", "body": "Sandboxing, least privilege, secret management, dependency review, and human inspection cannot be replaced by asking the same model to approve its output.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "practical-ai" ], "illustration": { "imageSearchTerm": "server rack security padlock", "imagePrompt": "A secure server rack behind a glass door fitted with a padlock and keycard reader, representing layered independent security controls.", "alt": "Security Needs Independent Controls", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Channel_in_a_Box_Server.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-test-secure-3.webp" }, "uid": "qhs1iag7qisq" }, { "id": "afe-p-test-secure-1", "shape": "pair", "sideA": { "modality": "text", "value": "Sandbox", "short": "Sandbox" }, "sideB": { "modality": "text", "value": "An isolated environment that limits a program's effects", "short": "An isolated environment that limits a program's effects" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "vocabulary" ], "uid": "1mohswhe80l2j" }, { "id": "afe-p-test-secure-2", "shape": "pair", "sideA": { "modality": "text", "value": "Secret", "short": "Secret" }, "sideB": { "modality": "text", "value": "Sensitive credential such as a token or private key", "short": "Sensitive credential such as a token or private key" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "vocabulary" ], "uid": "13mqyzk1mscjp0" }, { "id": "afe-d-test-secure", "shape": "definition", "term": { "modality": "text", "value": "Threat model", "short": "Threat model" }, "definition": { "modality": "text", "value": "A structured account of assets, attackers, entry points, and plausible harms.", "short": "A structured account of assets" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "definition" ], "uid": "1uwv885ouv92b" }, { "id": "afe-q-test-secure-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should unfamiliar generated code run?" }, "options": [ { "modality": "text", "value": "With every admin permission", "short": "Every admin permission" }, { "modality": "text", "value": "In a limited environment", "short": "Limited environment" }, { "modality": "text", "value": "Beside production secrets", "short": "By production secrets" }, { "modality": "text", "value": "Without reading the diff", "short": "Skip reading the diff" } ], "correctIndex": 1, "explanation": "Isolation limits damage while behavior is still uncertain.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "quiz" ], "uid": "14hmt1r137x6ft" }, { "id": "afe-q-test-secure-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which case tests access control?" }, "options": [ { "modality": "text", "value": "An empty optional field", "short": "Empty optional field" }, { "modality": "text", "value": "Unauthorized input", "short": "Unauthorized input" }, { "modality": "text", "value": "A large valid file", "short": "A large valid file" }, { "modality": "text", "value": "A slow network response", "short": "Slow network response" } ], "correctIndex": 1, "explanation": "An unauthorized case checks whether the system enforces access boundaries.", "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "quiz" ], "uid": "1cavd1p88bjcz" }, { "id": "afe-p-test-secure-3", "shape": "pair", "sideA": { "modality": "text", "value": "Dependency audit", "short": "Dependency audit" }, "sideB": { "modality": "text", "value": "A check of third-party packages and known risks", "short": "A check of third-party packages and known risks" }, "source": { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, "tags": [ "data-code", "test-secure", "vocabulary" ], "uid": "1lwuh7iy0rkim" }, { "id": "afe-f-verify-claims-1", "shape": "fact", "title": "Verify the Material Claims", "body": "Focus checking effort on statements that drive a decision, create risk, or would be costly to repeat if wrong.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Focus checking effort on statements that drive a decision, create risk, or would be costly to repeat if wrong.", "illustration": { "kind": "photo", "imageSearchTerm": "magnifying glass documents fact check", "imagePrompt": "Editorial documentary photograph illustrating fact checking sources citations. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A fact-checker comparing claims with original sources", "credit": "Pexels · Close-up of hand using magnifying glass to review documents. Ideal for financial themes.", "creditUrl": "https://www.pexels.com/photo/a-person-using-a-magnifying-glass-on-a-document-7821674/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-verify-claims-1.webp" }, "uid": "3zyti1vaol2i" }, { "id": "afe-f-verify-claims-2", "shape": "fact", "title": "Existence and Support Are Separate", "body": "A citation can be real while failing to support the attached claim, or support only a narrower qualified version.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "practical-ai" ], "illustration": { "imageSearchTerm": "bookmark wrong page book", "imagePrompt": "A real book on a shelf with a bookmark placed in it, the bookmark sitting several chapters away from the specific passage being referenced.", "alt": "Existence and Support Are Separate", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Screenshot_UC_Browser.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-verify-claims-2.webp" }, "uid": "1q8zi9ou398rc" }, { "id": "afe-f-verify-claims-3", "shape": "fact", "title": "Independent Evidence Matters", "body": "Rephrasing the question to the same model can be useful, but authoritative sources, calculations, and external tests provide stronger confirmation.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "practical-ai" ], "illustration": { "imageSearchTerm": "reference books calculator instrument", "imagePrompt": "Several independent reference books, a calculator, and a lab instrument arranged together around a single central question, representing multiple independent sources of confirmation.", "alt": "Independent Evidence Matters", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:How_WP_Works_books.JPG", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-verify-claims-3.webp" }, "uid": "t5vhdxa2hhyj" }, { "id": "afe-p-verify-claims-1", "shape": "pair", "sideA": { "modality": "text", "value": "Claim verification", "short": "Claim verification" }, "sideB": { "modality": "text", "value": "Checking a statement against appropriate evidence", "short": "Checking a statement against appropriate evidence" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "vocabulary" ], "uid": "inbgoj1d2yret" }, { "id": "afe-p-verify-claims-2", "shape": "pair", "sideA": { "modality": "text", "value": "Citation alignment", "short": "Citation alignment" }, "sideB": { "modality": "text", "value": "Whether a source actually supports the attached claim", "short": "Whether a source actually supports the attached claim" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "vocabulary" ], "uid": "1uakh5a1lw919m" }, { "id": "afe-d-verify-claims", "shape": "definition", "term": { "modality": "text", "value": "Corroboration", "short": "Corroboration" }, "definition": { "modality": "text", "value": "Independent evidence that supports or confirms a claim.", "short": "Independent evidence that supports or confirms a claim." }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "definition" ], "uid": "odbp7v1575t4l" }, { "id": "afe-q-verify-claims-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What must be checked about a citation?" }, "options": [ { "modality": "text", "value": "That it supports the claim", "short": "Supports the claim" }, { "modality": "text", "value": "That the title sounds relevant", "short": "Relevant title" }, { "modality": "text", "value": "That the answer is confident", "short": "Confident answer" }, { "modality": "text", "value": "That the URL looks long", "short": "Long-looking URL" } ], "correctIndex": 0, "explanation": "A source must be both genuine and aligned with the statement.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "quiz" ], "uid": "1ftul9hg0c3vn" }, { "id": "afe-q-verify-claims-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which check is most independent?" }, "options": [ { "modality": "text", "value": "Recalculate from source data", "short": "Recalculate at source" }, { "modality": "text", "value": "Request a longer rationale", "short": "Ask for more rationale" }, { "modality": "text", "value": "Repeat the same prompt", "short": "Repeat the same prompt" }, { "modality": "text", "value": "Ask for more confidence", "short": "Ask for confidence" } ], "correctIndex": 0, "explanation": "External inputs and arithmetic test the result beyond its wording.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "quiz" ], "uid": "oehfih1r7na27" }, { "id": "afe-cmp-verify-claims", "shape": "comparison", "prompt": "Which verification is stronger?", "correct": { "caption": "Open the cited source and compare the exact claim.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-verify-claims-open-cited-source.svg" }, "incorrect": { "caption": "Ask the model whether its own citation is probably correct.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-verify-claims-ask-model-whether.svg" }, "explanation": "Direct inspection tests both the source and its alignment.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "verify-claims", "comparison" ], "uid": "1sxd2e0s5erp0" }, { "id": "afe-f-privacy-1", "shape": "fact", "title": "Data Minimization Reduces Exposure", "body": "Provide only the fields and detail required for the task, and remove direct or indirect identifiers when they are unnecessary.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Provide only the fields and detail required for the task, and remove direct or indirect identifiers when they are unnecessary.", "illustration": { "kind": "photo", "imageSearchTerm": "padlock confidential document desk", "imagePrompt": "Editorial documentary photograph illustrating privacy redaction document. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A document with sensitive details carefully redacted", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Abloy-riippulukko.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-privacy-1.webp" }, "uid": "1wkjful1tsb21j" }, { "id": "afe-f-privacy-2", "shape": "fact", "title": "Policies Depend on Product and Setting", "body": "Retention, training use, human review, sharing, deletion, and administrator controls can vary across services and plans.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "practical-ai" ], "illustration": { "imageSearchTerm": "toggle switches settings row", "imagePrompt": "A row of several software-style toggle switches, some on and some off, representing varying policy configurations, no legible text.", "alt": "Policies Depend on Product and Setting", "depictable": false, "credit": "Pexels · Srattha Nualsate · Pexels License", "creditUrl": "https://www.pexels.com/photo/white-light-rocker-switches-on-gray-wall-12996907/", "subject": "two white plastic wall plates on a grey stone wall, the upper holding three rocker switches in a row and the lower two, all switches identical and unlabelled", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-privacy-2.webp" }, "uid": "1ksbqmybpi3qu" }, { "id": "afe-f-privacy-3", "shape": "fact", "title": "Other People's Data Is Not Yours to Volunteer", "body": "Private messages, records, images, and voices can affect people who did not choose to use the system.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "practical-ai" ], "illustration": { "imageSearchTerm": "locked photo album envelopes", "imagePrompt": "A closed photo album and a stack of sealed envelopes kept locked inside a glass case, representing other people's private information kept out of reach.", "alt": "Other People's Data Is Not Yours to Volunteer", "depictable": false, "credit": "Unsplash · Roman Kraft · Unsplash License", "creditUrl": "https://unsplash.com/photos/old-photos-in-brown-wooden-chest-7sPg5OLfExc", "subject": "An open wooden chest with a metal lock plate on its front, filled with old black-and-white portrait photographs and cabinet cards of unnamed people, a framed print behind.", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-privacy-3.webp" }, "uid": "1xarsdm19xvr4y" }, { "id": "afe-p-privacy-1", "shape": "pair", "sideA": { "modality": "text", "value": "Personal data", "short": "Personal data" }, "sideB": { "modality": "text", "value": "Information linked or linkable to a person", "short": "Information linked or linkable to a person" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "vocabulary" ], "uid": "uhs2ck15ialak" }, { "id": "afe-p-privacy-2", "shape": "pair", "sideA": { "modality": "text", "value": "Data minimization", "short": "Data minimization" }, "sideB": { "modality": "text", "value": "Limiting collection and sharing to what is necessary", "short": "Limiting collection and sharing to what is necessary" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "vocabulary" ], "uid": "1hl21071apj9px" }, { "id": "afe-d-privacy", "shape": "definition", "term": { "modality": "text", "value": "De-identification", "short": "De-identification" }, "definition": { "modality": "text", "value": "Reducing the link between data and the person it describes.", "short": "Reducing the link between data and the person it describes." }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "definition" ], "uid": "w0qidr1u8n0bp" }, { "id": "afe-q-privacy-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Best way to test a sensitive workflow?" }, "options": [ { "modality": "text", "value": "Disable access controls", "short": "Disable access control" }, { "modality": "text", "value": "Paste real customer records", "short": "Paste customer records" }, { "modality": "text", "value": "Use synthetic sample data", "short": "Synthetic sample data" }, { "modality": "text", "value": "Share every identifier", "short": "Share every identifier" } ], "correctIndex": 2, "explanation": "Synthetic examples reveal workflow needs without exposing real people.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "quiz" ], "uid": "p0isei6ht652" }, { "id": "afe-q-privacy-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should be checked before upload?" }, "options": [ { "modality": "text", "value": "Current data-handling terms", "short": "Current data terms" }, { "modality": "text", "value": "A marketing slogan", "short": "A marketing slogan" }, { "modality": "text", "value": "Only the interface color", "short": "Interface color only" }, { "modality": "text", "value": "The model's tone", "short": "The model's tone" } ], "correctIndex": 0, "explanation": "Retention and use depend on the actual product, plan, and settings.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "quiz" ], "uid": "1lqlrtxo3t94r" }, { "id": "afe-proc-privacy", "shape": "procedure", "goal": "Minimize data before using an AI tool", "steps": [ "Identify the task's essential fields", "Remove names and unnecessary identifiers", "Replace real values with placeholders when possible", "Confirm authorization, policy, and account settings", "Delete temporary inputs and outputs according to policy" ], "notes": "De-identification can fail when remaining details still point to a person.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "privacy", "procedure" ], "uid": "1264hcp1yeii0n" }, { "id": "afe-f-manipulation-1", "shape": "fact", "title": "Retrieved Content Is Untrusted", "body": "Text found in a file or webpage can contain manipulative instructions and should not outrank the workflow's trusted controls.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Text found in a file or webpage can contain manipulative instructions and should not outrank the workflow's trusted controls.", "illustration": { "kind": "photo", "imageSearchTerm": "phishing email warning laptop", "imagePrompt": "A laptop screen displaying a suspicious email with a warning triangle icon beside it, in a dim home office, no legible text.", "alt": "A suspicious message being inspected for manipulation", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Fishingwebseite_deutschebank.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-manipulation-1.webp" }, "uid": "xqv0ax3bfewj" }, { "id": "afe-f-manipulation-2", "shape": "fact", "title": "Familiar Sound Is Not Authentication", "body": "Voice cloning can imitate a person well enough to support an emergency or payment scam.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "practical-ai" ], "illustration": { "imageSearchTerm": "phone call voice scam", "imagePrompt": "A person looking concerned while holding a phone to their ear, with a faint abstract sound-wave glow near the speaker, suggesting a suspicious call.", "alt": "Familiar Sound Is Not Authentication", "depictable": true }, "uid": "kkj2omf7utgi" }, { "id": "afe-f-manipulation-3", "shape": "fact", "title": "Verify through a Trusted Channel", "body": "Use a known phone number, account, contact, or agreed code rather than contact information supplied in the suspicious message.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "practical-ai" ], "illustration": { "imageSearchTerm": "phone contacts list call", "imagePrompt": "A hand scrolling through a trusted contacts list on a phone to place a call, instead of using a number from a separate suspicious message on another screen, no legible text.", "alt": "Verify through a Trusted Channel", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Groningen_Vismarkt_12.JPG", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-manipulation-3.webp" }, "uid": "dpoitujdswfi" }, { "id": "afe-p-manipulation-1", "shape": "pair", "sideA": { "modality": "text", "value": "Prompt injection", "short": "Prompt injection" }, "sideB": { "modality": "text", "value": "Untrusted text that tries to redirect an AI workflow", "short": "Untrusted text that tries to redirect an AI workflow" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "vocabulary" ], "uid": "6wf3rkfpwgwk" }, { "id": "afe-p-manipulation-2", "shape": "pair", "sideA": { "modality": "text", "value": "Out-of-band check", "short": "Out-of-band check" }, "sideB": { "modality": "text", "value": "Verification through a separate trusted channel", "short": "Verification through a separate trusted channel" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "vocabulary" ], "uid": "w9nc2x11despf" }, { "id": "afe-d-manipulation", "shape": "definition", "term": { "modality": "text", "value": "Social engineering", "short": "Social engineering" }, "definition": { "modality": "text", "value": "Manipulating a person into revealing information or taking an unsafe action.", "short": "Manipulating a person into revealing information or taking an unsafe action." }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "definition" ], "uid": "1y76nl58q5gg7" }, { "id": "afe-q-manipulation-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "A familiar voice requests urgent money. First move?" }, "options": [ { "modality": "text", "value": "Share an account password", "short": "Share account password" }, { "modality": "text", "value": "Pay before the call ends", "short": "Pay before call ends" }, { "modality": "text", "value": "Trust the caller ID", "short": "Trust the caller ID" }, { "modality": "text", "value": "Verify through a known channel", "short": "Verify known channel" } ], "correctIndex": 3, "explanation": "Voice and caller ID can be spoofed; independent contact reduces risk.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "quiz" ], "uid": "tawodvf6yhq1" }, { "id": "afe-q-manipulation-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "How should a webpage instruction affect an agent?" }, "options": [ { "modality": "text", "value": "Grant new permissions", "short": "Grant new permissions" }, { "modality": "text", "value": "Disable the logs", "short": "Disable the logs" }, { "modality": "text", "value": "Override the user", "short": "Override the user" }, { "modality": "text", "value": "Treat it as untrusted content", "short": "Treat as untrusted" } ], "correctIndex": 3, "explanation": "Retrieved text is data and may be hostile.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "quiz" ], "uid": "1hqhgwph2uyr7" }, { "id": "afe-p-manipulation-3", "shape": "pair", "sideA": { "modality": "text", "value": "Urgency cue", "short": "Urgency cue" }, "sideB": { "modality": "text", "value": "Pressure designed to prevent careful verification", "short": "Pressure designed to prevent careful verification" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "manipulation", "vocabulary" ], "uid": "vf5gvp1vvziy3" }, { "id": "afe-f-high-stakes-1", "shape": "fact", "title": "High Stakes Change the Standard", "body": "The greater the possible harm, the stronger the evidence, expertise, privacy, review, and accountability that are required.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "The greater the possible harm, the stronger the evidence, expertise, privacy, review, and accountability that are required.", "illustration": { "kind": "photo", "imageSearchTerm": "doctor talking patient clinic", "imagePrompt": "Editorial documentary photograph illustrating professional consultation human review. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A person consulting a qualified professional with organized notes", "credit": "Pexels · Doctor discusses health records with a patient in a clinical setting.", "creditUrl": "https://www.pexels.com/photo/male-doctor-talking-to-a-patient-6129444/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-high-stakes-1.webp" }, "uid": "g20jzf57w14h" }, { "id": "afe-f-high-stakes-2", "shape": "fact", "title": "Preparation Is a Safer Role", "body": "AI can help assemble questions, timelines, and source excerpts while a qualified person retains diagnosis, advice, or decision authority.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "practical-ai" ], "illustration": { "imageSearchTerm": "professional reviewing prepared folder", "imagePrompt": "A professional in an office reviewing a neatly prepared folder of timelines and source excerpts handed to them, before making the final decision themselves.", "alt": "Preparation Is a Safer Role", "depictable": false, "credit": "Pexels · A professional woman reviewing documents at her office desk with a laptop and files.", "creditUrl": "https://www.pexels.com/photo/woman-holding-a-brown-folder-while-working-in-the-office-7654450/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-high-stakes-2.webp" }, "uid": "ufqk1m1xvupti" }, { "id": "afe-f-high-stakes-3", "shape": "fact", "title": "Affected People Need Recourse", "body": "When an AI-assisted decision changes access, opportunity, or rights, people need notice, human review, and a correction or appeal path.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "practical-ai" ], "illustration": { "imageSearchTerm": "appeal form help window", "imagePrompt": "A person at a help window handing over an appeal form to a clerk, representing a path to challenge a decision.", "alt": "Affected People Need Recourse", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Rosa_sp.99.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-high-stakes-3.webp" }, "uid": "10093gz4e1ya9" }, { "id": "afe-p-high-stakes-1", "shape": "pair", "sideA": { "modality": "text", "value": "Decision support", "short": "Decision support" }, "sideB": { "modality": "text", "value": "Information that helps a responsible person decide", "short": "Information that helps a responsible person decide" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "vocabulary" ], "uid": "1hajyg1py6ilr" }, { "id": "afe-p-high-stakes-2", "shape": "pair", "sideA": { "modality": "text", "value": "Decision authority", "short": "Decision authority" }, "sideB": { "modality": "text", "value": "Power and accountability to make the final choice", "short": "Power and accountability to make the final choice" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "vocabulary" ], "uid": "p37sck11nwu3o" }, { "id": "afe-d-high-stakes", "shape": "definition", "term": { "modality": "text", "value": "High-stakes use", "short": "High-stakes use" }, "definition": { "modality": "text", "value": "An application where errors can materially affect safety, rights, health, livelihood, or access.", "short": "An application where errors can materially affect safety" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "definition" ], "uid": "cl0utz162lnwh" }, { "id": "afe-q-high-stakes-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is a safer medical use?" }, "options": [ { "modality": "text", "value": "Prepare questions for a clinician", "short": "Clinician questions" }, { "modality": "text", "value": "Hide symptoms", "short": "Hide symptoms" }, { "modality": "text", "value": "Change medication alone", "short": "Change meds alone" }, { "modality": "text", "value": "Replace emergency care", "short": "Replace emergency care" } ], "correctIndex": 0, "explanation": "Preparation supports the human consultation without claiming diagnosis authority.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "quiz" ], "uid": "175my5g1nthmec" }, { "id": "afe-q-high-stakes-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should an affected person receive?" }, "options": [ { "modality": "text", "value": "No explanation", "short": "No explanation" }, { "modality": "text", "value": "A review and correction path", "short": "Review + correction" }, { "modality": "text", "value": "A secret score only", "short": "A secret score only" }, { "modality": "text", "value": "Automatic finality", "short": "Automatic finality" } ], "correctIndex": 1, "explanation": "Consequential systems need accountable recourse.", "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "quiz" ], "uid": "1ar1841j8rybk" }, { "id": "afe-p-high-stakes-3", "shape": "pair", "sideA": { "modality": "text", "value": "Qualified reviewer", "short": "Qualified reviewer" }, "sideB": { "modality": "text", "value": "A person with relevant expertise and accountability", "short": "A person with relevant expertise and accountability" }, "source": { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, "tags": [ "verify-protect", "high-stakes", "vocabulary" ], "uid": "b57ioxckih8" }, { "id": "afe-f-choose-tools-1", "shape": "fact", "title": "Task Fit Comes Before Popularity", "body": "The best tool is the one that meets the real input, output, quality, control, and workflow requirements—not necessarily the most famous model.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "The best tool is the one that meets the real input, output, quality, control, and workflow requirements—not necessarily the most famous model.", "illustration": { "kind": "photo", "imageSearchTerm": "toolbox choosing right tool", "imagePrompt": "Editorial documentary photograph illustrating toolbox choosing right tool. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A carefully chosen tool beside a varied toolbox", "credit": "Pexels · From above of crop anonymous skillful male in casual clothes choosing screwdriver from toolbox in daylight", "creditUrl": "https://www.pexels.com/photo/faceless-master-choosing-screwdriver-from-toolbox-4792511/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-choose-tools-1.webp" }, "uid": "1t9zckw1dufi64" }, { "id": "afe-f-choose-tools-2", "shape": "fact", "title": "Capabilities Include Controls", "body": "Permissions, logs, deletion, export, administrator settings, and approval gates are part of a tool's practical capability.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "practical-ai" ], "illustration": { "imageSearchTerm": "control panel switches dials", "imagePrompt": "A control panel with several distinct switches, dials, and an export tray, representing a tool's full range of practical controls, no legible text.", "alt": "Capabilities Include Controls", "depictable": false, "credit": "Pexels · Detailed view of switches and dials on an industrial control panel, showcasing engineering complexity.", "creditUrl": "https://www.pexels.com/photo/close-up-of-the-switches-of-a-machine-13401910/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-choose-tools-2.webp" }, "uid": "1ncpruy3prr0e" }, { "id": "afe-f-choose-tools-3", "shape": "fact", "title": "Run Representative Tests", "body": "A short evaluation set from your real work reveals quality, speed, failure patterns, and cost better than a polished demo.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "practical-ai" ], "illustration": { "imageSearchTerm": "real work samples testing", "imagePrompt": "A small stack of real, worn work samples being timed and checked on a desk, set apart from a single glossy polished demo sample.", "alt": "Run Representative Tests", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Adrian-_molina_2013_001.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-choose-tools-3.webp" }, "uid": "mx5ovl4oic2j" }, { "id": "afe-p-choose-tools-1", "shape": "pair", "sideA": { "modality": "text", "value": "Task fit", "short": "Task fit" }, "sideB": { "modality": "text", "value": "How well a tool matches a defined job", "short": "How well a tool matches a defined job" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "vocabulary" ], "uid": "tpqb9syvcpak" }, { "id": "afe-p-choose-tools-2", "shape": "pair", "sideA": { "modality": "text", "value": "Exit option", "short": "Exit option" }, "sideB": { "modality": "text", "value": "A way to export data or stop using the service", "short": "A way to export data or stop using the service" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "vocabulary" ], "uid": "jngcuwujbj4" }, { "id": "afe-d-choose-tools", "shape": "definition", "term": { "modality": "text", "value": "Tool evaluation", "short": "Tool evaluation" }, "definition": { "modality": "text", "value": "A structured comparison of products against task, quality, risk, control, and cost criteria.", "short": "A structured comparison of products against task" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "definition" ], "uid": "1648wez1tvt989" }, { "id": "afe-q-choose-tools-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should lead tool selection?" }, "options": [ { "modality": "text", "value": "A defined task and test set", "short": "Defined task + tests" }, { "modality": "text", "value": "The longest feature page", "short": "Longest feature page" }, { "modality": "text", "value": "A viral demo", "short": "A viral demo" }, { "modality": "text", "value": "A mascot", "short": "A mascot" } ], "correctIndex": 0, "explanation": "Requirements and representative cases reveal fit.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "quiz" ], "uid": "1kogytrbize4d" }, { "id": "afe-q-choose-tools-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Which feature affects trust boundaries?" }, "options": [ { "modality": "text", "value": "Button color", "short": "Button color" }, { "modality": "text", "value": "Brand popularity", "short": "Brand popularity" }, { "modality": "text", "value": "Data retention controls", "short": "Data retention control" }, { "modality": "text", "value": "A playful name", "short": "A playful name" } ], "correctIndex": 2, "explanation": "Retention determines how submitted information may persist.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "quiz" ], "uid": "5z9m7d1exuayv" }, { "id": "afe-cmp-choose-tools", "shape": "comparison", "prompt": "Which selection process is more durable?", "correct": { "caption": "Test real cases and compare quality, controls, and cost.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-choose-tools-test-real-cases.svg" }, "incorrect": { "caption": "Choose the tool with the most exciting launch video.", "imageUrl": "/packs/ai-for-everyone/diagrams/afe-cmp-choose-tools-choose-tool-exciting.svg" }, "explanation": "Representative evidence is more informative than promotional polish.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "choose-tools", "comparison" ], "uid": "fso37ylcrbz6" }, { "id": "afe-f-templates-1", "shape": "fact", "title": "Templates Separate Stable and Variable Parts", "body": "Purpose, process, and format may stay fixed while audience, source text, dates, and options change each run.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "Purpose, process, and format may stay fixed while audience, source text, dates, and options change each run.", "illustration": { "kind": "photo", "imageSearchTerm": "blank business card template mockup", "imagePrompt": "Editorial documentary photograph illustrating reusable prompt template cards. Natural light, believable contemporary setting, clear subject, generous card-safe composition, no readable text, no logos, no futuristic holograms.", "alt": "A reusable prompt template with clearly marked fields", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Beth_Phoenix_and_Kelly_Kelly_2.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-templates-1.webp" }, "uid": "aure4h175obnr" }, { "id": "afe-f-templates-2", "shape": "fact", "title": "Placeholders Need Instructions", "body": "A label such as {SOURCE_TEXT} should say what belongs there, acceptable length, and what to do when the input is missing.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "practical-ai" ], "illustration": { "imageSearchTerm": "blank form sticky note", "imagePrompt": "A form with an empty highlighted blank space and a small instructional sticky note attached beside it, no legible text.", "alt": "Placeholders Need Instructions", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Podac%C3%AD_l%C3%ADstek.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-templates-2.webp" }, "uid": "1lax0rd1acke0v" }, { "id": "afe-f-templates-3", "shape": "fact", "title": "Versioning Supports Learning", "body": "A change note and test results preserve why a template evolved and make regressions easier to diagnose.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "practical-ai" ], "illustration": { "imageSearchTerm": "stacked revision documents sequence", "imagePrompt": "A stack of numbered document revisions fanned out in sequence on a desk, each slightly different from the last, no legible text.", "alt": "Versioning Supports Learning", "depictable": false, "credit": "Pexels · Close-up of stacked binders filled with documents for office or educational use.", "creditUrl": "https://www.pexels.com/photo/pile-of-folders-357514/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-templates-3.webp" }, "uid": "1y2x6x810y1zrs" }, { "id": "afe-p-templates-1", "shape": "pair", "sideA": { "modality": "text", "value": "Template variable", "short": "Template variable" }, "sideB": { "modality": "text", "value": "A field replaced for each run", "short": "A field replaced for each run" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "vocabulary" ], "uid": "1vf4wlxgkldxb" }, { "id": "afe-p-templates-2", "shape": "pair", "sideA": { "modality": "text", "value": "Stable instruction", "short": "Stable instruction" }, "sideB": { "modality": "text", "value": "A rule intended to remain across runs", "short": "A rule intended to remain across runs" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "vocabulary" ], "uid": "ye1j2w19cb6zs" }, { "id": "afe-d-templates", "shape": "definition", "term": { "modality": "text", "value": "Prompt template", "short": "Prompt template" }, "definition": { "modality": "text", "value": "A reusable prompt structure with named variables and consistent instructions.", "short": "A reusable prompt structure with named variables and consistent instructions." }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "definition" ], "uid": "m7ppi41ggj214" }, { "id": "afe-q-templates-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What belongs in a reusable template?" }, "options": [ { "modality": "text", "value": "Unexplained copied chat", "short": "Unexplained chat copy" }, { "modality": "text", "value": "Private data by default", "short": "Use private data first" }, { "modality": "text", "value": "A hidden final answer", "short": "A hidden final answer" }, { "modality": "text", "value": "Named input placeholders", "short": "Named input fields" } ], "correctIndex": 3, "explanation": "Named fields make the structure reusable and inspectable.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "quiz" ], "uid": "1i1ok6y157ghtm" }, { "id": "afe-q-templates-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why version a template?" }, "options": [ { "modality": "text", "value": "Track changes and regressions", "short": "Track regressions" }, { "modality": "text", "value": "Hide the owner", "short": "Hide the owner" }, { "modality": "text", "value": "Avoid all testing", "short": "Avoid all testing" }, { "modality": "text", "value": "Guarantee one perfect output", "short": "Guarantee perfection" } ], "correctIndex": 0, "explanation": "Versions connect prompt changes to observed performance.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "quiz" ], "uid": "1cblqoe1ew4yky" }, { "id": "afe-p-templates-3", "shape": "pair", "sideA": { "modality": "text", "value": "Template owner", "short": "Template owner" }, "sideB": { "modality": "text", "value": "The person responsible for review and updates", "short": "The person responsible for review and updates" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "templates", "vocabulary" ], "uid": "151jf4gj8ad08" }, { "id": "afe-f-evaluate-1", "shape": "fact", "title": "Representative Cases Beat One Demo", "body": "A small set spanning normal, edge, and failure conditions provides a more stable signal than one memorable success.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A small set spanning normal, edge, and failure conditions provides a more stable signal than one memorable success.", "illustration": { "kind": "photo", "imageSearchTerm": "evaluation scorecard test cases", "imagePrompt": "A scorecard on a clipboard with checkmarks across a varied row of small sample items representing normal, edge, and failure conditions, no legible text.", "alt": "A scorecard comparing outputs across representative test cases", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Abstimmungsger%C3%A4t_Reply_IQ_Fleetwood_Group.JPG", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-evaluate-1.webp" }, "uid": "19fgvdy138hdxm" }, { "id": "afe-f-evaluate-2", "shape": "fact", "title": "Severity Matters beside Frequency", "body": "A rare harmful error may deserve more attention than many harmless formatting misses.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "practical-ai" ], "illustration": { "imageSearchTerm": "red flag among gray", "imagePrompt": "A single small red warning flag standing out among a large field of many small gray flags, representing one severe outlier among frequent minor ones.", "alt": "Severity Matters beside Frequency", "depictable": false, "credit": "Pexels · A red flag on a lifeguard tower set against a cloudy gray sky, symbolizing caution.", "creditUrl": "https://www.pexels.com/photo/red-flag-on-pole-swaying-by-the-wind-4996781/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-evaluate-2.webp" }, "uid": "1gm47x01261104" }, { "id": "afe-f-evaluate-3", "shape": "fact", "title": "Changes Require Retesting", "body": "New prompts, models, tools, integrations, or data can alter performance and revive old failures.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "practical-ai" ], "illustration": { "imageSearchTerm": "test track vehicle rerun", "imagePrompt": "A test track being run a second time after one visible component of the vehicle has been swapped out for a different part.", "alt": "Changes Require Retesting", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:CHARLE_DE_GAULLE.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-evaluate-3.webp" }, "uid": "1lccbyc80gms" }, { "id": "afe-p-evaluate-1", "shape": "pair", "sideA": { "modality": "text", "value": "Evaluation case", "short": "Evaluation case" }, "sideB": { "modality": "text", "value": "An input with expected properties used for testing", "short": "An input with expected properties used for testing" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "vocabulary" ], "uid": "1ko5i7b1es7lit" }, { "id": "afe-p-evaluate-2", "shape": "pair", "sideA": { "modality": "text", "value": "Stop-ship failure", "short": "Stop-ship failure" }, "sideB": { "modality": "text", "value": "A defect serious enough to block release", "short": "A defect serious enough to block release" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "vocabulary" ], "uid": "waxtt1bnh4r7" }, { "id": "afe-d-evaluate", "shape": "definition", "term": { "modality": "text", "value": "Evaluation rubric", "short": "Evaluation rubric" }, "definition": { "modality": "text", "value": "A set of dimensions and standards used to score outputs consistently.", "short": "A set of dimensions and standards used to score outputs consistently." }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "definition" ], "uid": "kne5kx1vnssxr" }, { "id": "afe-q-evaluate-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What should an evaluation set include?" }, "options": [ { "modality": "text", "value": "Promotional examples", "short": "Promotional examples" }, { "modality": "text", "value": "No expected behavior", "short": "No expected behavior" }, { "modality": "text", "value": "Only the easiest success", "short": "Only easiest success" }, { "modality": "text", "value": "Normal and edge cases", "short": "Normal and edge cases" } ], "correctIndex": 3, "explanation": "Varied cases expose both routine quality and failure boundaries.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "quiz" ], "uid": "16d4811kmkauz" }, { "id": "afe-q-evaluate-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Why track severity?" }, "options": [ { "modality": "text", "value": "Averages show every risk", "short": "Averages show all risk" }, { "modality": "text", "value": "Every typo is equally harmful", "short": "Typos equally risky" }, { "modality": "text", "value": "Cost never matters", "short": "Cost never matters" }, { "modality": "text", "value": "Rare errors can be catastrophic", "short": "Rare errors: fatal" } ], "correctIndex": 3, "explanation": "Impact changes how a failure should be prioritized.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "quiz" ], "uid": "101anjhuv50dn" }, { "id": "afe-p-evaluate-3", "shape": "pair", "sideA": { "modality": "text", "value": "Regression case", "short": "Regression case" }, "sideB": { "modality": "text", "value": "An example retained to prevent a known failure from returning", "short": "An example retained to prevent a known failure from returning" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "evaluate", "vocabulary" ], "uid": "1midd881wkyts0" }, { "id": "afe-f-handoffs-1", "shape": "fact", "title": "Start with Reversible Work", "body": "A bounded task such as outlining, reformatting, or practice questions lets you learn without giving the system costly authority.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "practical-ai" ], "factVariant": "image-heavy", "imageCaption": "A bounded task such as outlining, reformatting, or practice questions lets you learn without giving the system costly authority.", "illustration": { "kind": "photo", "imageSearchTerm": "practice calendar roadmap planning", "imagePrompt": "A desk calendar with a few early low-risk practice tasks checked off, next to a simple hand-drawn roadmap sketch, no legible text.", "alt": "A learning roadmap beside a short practice calendar", "depictable": true, "credit": "Unsplash · Towfiqu barbhuiya · Unsplash License", "creditUrl": "https://unsplash.com/photos/a-calendar-with-red-push-buttons-pinned-to-it-bwOAixLG0uc", "subject": "A white monthly wall calendar (Sun–Fri columns) with four red push pins on a diagonal run of dates and the 30th circled in red marker; no logos or words beyond day names and numbers", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-handoffs-1.webp" }, "uid": "12zf8rf1en94f5" }, { "id": "afe-f-handoffs-2", "shape": "fact", "title": "Measure Corrections, Not Just Speed", "body": "Time saved before review can disappear when errors, rewriting, privacy cleanup, or rework are counted.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "practical-ai" ], "illustration": { "imageSearchTerm": "stopwatch corrected pages pile", "imagePrompt": "A stopped stopwatch resting next to a separate pile of red-pen corrected pages, showing the fast time didn't include the cleanup work.", "alt": "Measure Corrections, Not Just Speed", "depictable": false, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Atemschutz.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/afe-f-handoffs-2.webp" }, "uid": "5d6i7zbhhs75" }, { "id": "afe-p-handoffs-1", "shape": "pair", "sideA": { "modality": "text", "value": "Bounded pilot", "short": "Bounded pilot" }, "sideB": { "modality": "text", "value": "A limited trial with success and stop criteria", "short": "A limited trial with success and stop criteria" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "vocabulary" ], "uid": "6u2i3i14xsns2" }, { "id": "afe-p-handoffs-2", "shape": "pair", "sideA": { "modality": "text", "value": "Workflow handoff", "short": "Workflow handoff" }, "sideB": { "modality": "text", "value": "A deliberate move to the pack that owns deeper scope", "short": "A deliberate move to the pack that owns deeper scope" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "vocabulary" ], "uid": "1lx2h0n1a1v3hl" }, { "id": "afe-d-handoffs", "shape": "definition", "term": { "modality": "text", "value": "Reversible workflow", "short": "Reversible workflow" }, "definition": { "modality": "text", "value": "A process whose outputs or changes can be safely reviewed and undone.", "short": "A process whose outputs or changes can be safely reviewed and undone." }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "definition" ], "uid": "1guaxk7158dswt" }, { "id": "afe-q-handoffs-1", "shape": "mcqShort", "prompt": { "modality": "text", "value": "What is a good first personal pilot?" }, "options": [ { "modality": "text", "value": "Publishing private records", "short": "Publish private data" }, { "modality": "text", "value": "An unsupervised legal decision", "short": "Solo legal decision" }, { "modality": "text", "value": "Automatic payments", "short": "Automatic payments" }, { "modality": "text", "value": "A reversible recurring task", "short": "Reversible repeat task" } ], "correctIndex": 3, "explanation": "Low-consequence repetition supports learning and measurement.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "quiz" ], "uid": "p1cl4s1kbqm4c" }, { "id": "afe-q-handoffs-2", "shape": "mcqShort", "prompt": { "modality": "text", "value": "Where do transformer mechanics belong?" }, "options": [ { "modality": "text", "value": "LLM 101", "short": "LLM 101" }, { "modality": "text", "value": "The Infinity Machine", "short": "The Infinity Machine" }, { "modality": "text", "value": "AI Start Here", "short": "AI Start Here" }, { "modality": "text", "value": "AI for Small Business", "short": "AI for Small Business" } ], "correctIndex": 0, "explanation": "LLM 101 owns the deeper model-mechanics scope.", "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "quiz" ], "uid": "40dpb21rrj6hi" }, { "id": "afe-p-handoffs-3", "shape": "pair", "sideA": { "modality": "text", "value": "AI for Small Business", "short": "AI for Small Business" }, "sideB": { "modality": "text", "value": "Team workflows, governance, and measured business pilots", "short": "Team workflows" }, "source": { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, "tags": [ "personal-workflow", "handoffs", "vocabulary" ], "uid": "1g5otas2rjqe" }, { "id": "czr-ai-for-everyone-afe-d-give-context", "shape": "cloze", "tags": [ "prompt-anatomy", "give-context", "definition" ], "template": "___ — Information supplied with a request that helps determine a relevant response", "answer": "Context", "distractors": [ "SOURCE TEXT", "Task specification", "Output format" ], "derivedFrom": "afe-d-give-context", "explanation": "Compare: Task specification — A concise statement of the action, subject, and intended result. Output format — The structure or representation in which a response should be delivered.", "uid": "192f3vs170e09g" }, { "id": "czr-ai-for-everyone-afe-d-define-success", "shape": "cloze", "tags": [ "prompt-anatomy", "define-success", "definition" ], "template": "___ — A set of criteria used to judge the quality of an output", "answer": "Rubric", "distractors": [ "Output format", "Task specification", "Context" ], "derivedFrom": "afe-d-define-success", "explanation": "Compare: Output format — The structure or representation in which a response should be delivered. Task specification — A concise statement of the action, subject, and intended result.", "uid": "64f7i1in73oe" }, { "id": "czr-ai-for-everyone-afe-d-one-shot", "shape": "cloze", "tags": [ "learn-by-showing", "one-shot", "definition" ], "template": "___ — A worked example included in a prompt to show how an input should map to an output", "answer": "Demonstration", "distractors": [ "Prompt iteration", "Few-shot prompting", "EXAMPLE OUTPUT" ], "derivedFrom": "afe-d-one-shot", "explanation": "Compare: Prompt iteration — A cycle of testing, diagnosing, revising, and comparing instructions. Few-shot prompting — Prompting with a small set of worked demonstrations.", "uid": "mtmhxx17dmqnr" }, { "id": "czr-ai-for-everyone-afe-d-iterate", "shape": "cloze", "tags": [ "learn-by-showing", "iterate", "definition" ], "template": "___ — A cycle of testing, diagnosing, revising, and comparing instructions", "answer": "Prompt iteration", "distractors": [ "Zero-shot prompting", "Demonstration", "Few-shot prompting" ], "derivedFrom": "afe-d-iterate", "explanation": "Compare: Zero-shot prompting — Requesting a task without supplying a completed demonstration. Demonstration — A worked example included in a prompt to show how an input should map to an output.", "uid": "1j37vyf179kq6l" }, { "id": "czr-ai-for-everyone-afe-d-visible-reasons", "shape": "cloze", "tags": [ "roles-reasoning", "visible-reasons", "definition" ], "template": "___ — A stated justification for a conclusion or choice", "answer": "Rationale", "distractors": [ "Persona prompting", "Task decomposition", "Calibration" ], "derivedFrom": "afe-d-visible-reasons", "explanation": "Compare: Persona prompting — Directing a model to respond from a named role or perspective. Task decomposition — Dividing a complex job into smaller, inspectable stages.", "uid": "1yj7v4o1lrp7hw" }, { "id": "czr-ai-for-everyone-afe-d-surface-uncertainty", "shape": "cloze", "tags": [ "roles-reasoning", "surface-uncertainty", "definition" ], "template": "___ — How closely stated confidence corresponds to observed accuracy", "answer": "Calibration", "distractors": [ "Task decomposition", "Rationale", "Persona prompting" ], "derivedFrom": "afe-d-surface-uncertainty", "explanation": "Compare: Task decomposition — Dividing a complex job into smaller, inspectable stages. Rationale — A stated justification for a conclusion or choice.", "uid": "1v96y8kqlp3tg" }, { "id": "czr-ai-for-everyone-afe-d-tools-permissions", "shape": "cloze", "tags": [ "agents", "tools-permissions", "definition" ], "template": "___ — Giving a user or system only the minimum access needed for its task", "answer": "Least privilege", "distractors": [ "Human-in-the-loop", "Prompt injection", "AI agent" ], "derivedFrom": "afe-d-tools-permissions", "explanation": "Compare: Human-in-the-loop — A workflow in which a person reviews, directs, or approves selected system actions. Prompt injection — Untrusted content designed to override or redirect an AI system's instructions.", "uid": "8wfb3evzxbju" }, { "id": "czr-ai-for-everyone-afe-d-agent-failures", "shape": "cloze", "tags": [ "agents", "agent-failures", "definition" ], "template": "___ — Untrusted content designed to override or redirect an AI system's instructions", "answer": "Prompt injection", "distractors": [ "AI agent", "Human-in-the-loop", "Least privilege" ], "derivedFrom": "afe-d-agent-failures", "explanation": "Compare: AI agent — A goal-directed system that can choose and execute actions across multiple steps. Human-in-the-loop — A workflow in which a person reviews, directs, or approves selected system actions.", "uid": "1k4hdp2esp7lm" }, { "id": "czr-ai-for-everyone-afe-d-video-planning", "shape": "cloze", "tags": [ "multimodal", "video-planning", "definition" ], "template": "___ — A sequence of images or panels that plans how a video will unfold", "answer": "Storyboard", "distractors": [ "Creative brief", "An ordered plan of camera setups", "Consistency of visual details across shots" ], "derivedFrom": "afe-d-video-planning", "explanation": "Compare: Creative brief — A concise statement of a visual project's goal, subject, audience, and constraints. An ordered plan of camera setups — Shot list.", "uid": "fmbn75sknjof" }, { "id": "czr-ai-for-everyone-afe-d-summarize", "shape": "cloze", "tags": [ "writing-documents", "summarize", "definition" ], "template": "___ — A condensed restatement written in new language rather than copied passages", "answer": "Abstractive summary", "distractors": [ "Tone", "Document grounding", "Revision brief" ], "derivedFrom": "afe-d-summarize", "explanation": "Compare: Tone — The attitude and relationship conveyed by choices of wording, rhythm, and emphasis. Document grounding — Constraining an answer to information found in a supplied set of documents.", "uid": "vrjh38gpnfvk" }, { "id": "czr-ai-for-everyone-afe-d-document-qa", "shape": "cloze", "tags": [ "writing-documents", "document-qa", "definition" ], "template": "___ — Constraining an answer to information found in a supplied set of documents", "answer": "Document grounding", "distractors": [ "Revision brief", "Tone", "Abstractive summary" ], "derivedFrom": "afe-d-document-qa", "explanation": "Compare: Revision brief — Instructions describing what to change, what to preserve, and how the result will be judged. Tone — The attitude and relationship conveyed by choices of wording, rhythm, and emphasis.", "uid": "1isf9nehf8p5i" }, { "id": "czr-ai-for-everyone-afe-d-tutor", "shape": "cloze", "tags": [ "research-learning", "tutor", "definition" ], "template": "___ — Information intended to improve learning during practice rather than only score the final result", "answer": "Formative feedback", "distractors": [ "Evidence table", "Decision matrix", "A hint that preserves the learner's next step" ], "derivedFrom": "afe-d-tutor", "explanation": "Compare: Evidence table — A structured record linking claims to sources, passages, dates, and confidence notes. Decision matrix — A table that compares options across explicitly chosen criteria.", "uid": "1gj1qz1aacurb" }, { "id": "czr-ai-for-everyone-afe-d-prepare-data", "shape": "cloze", "tags": [ "data-code", "prepare-data", "definition" ], "template": "___ — A structured description of fields, types, relationships, and constraints", "answer": "Schema", "distractors": [ "Threat model", "Data validation", "Code review" ], "derivedFrom": "afe-d-prepare-data", "explanation": "Compare: Threat model — A structured account of assets, attackers, entry points, and plausible harms. Data validation — Rules that restrict or check values entered into a field or cell.", "uid": "xrp9405aak00" }, { "id": "czr-ai-for-everyone-afe-d-coding-cycle", "shape": "cloze", "tags": [ "data-code", "coding-cycle", "definition" ], "template": "___ — Systematic inspection of a change for correctness, clarity, security, and fit", "answer": "Code review", "distractors": [ "Schema", "Threat model", "Data validation" ], "derivedFrom": "afe-d-coding-cycle", "explanation": "Compare: Schema — A structured description of fields, types, relationships, and constraints. Threat model — A structured account of assets, attackers, entry points, and plausible harms.", "uid": "oyp0wj1939bkx" }, { "id": "czr-ai-for-everyone-afe-d-verify-claims", "shape": "cloze", "tags": [ "verify-protect", "verify-claims", "definition" ], "template": "___ — Independent evidence that supports or confirms a claim", "answer": "Corroboration", "distractors": [ "Social engineering", "De-identification", "High-stakes use" ], "derivedFrom": "afe-d-verify-claims", "explanation": "Compare: Social engineering — Manipulating a person into revealing information or taking an unsafe action. De-identification — Reducing the link between data and the person it describes.", "uid": "usihk414l80jo" }, { "id": "czr-ai-for-everyone-afe-d-manipulation", "shape": "cloze", "tags": [ "verify-protect", "manipulation", "definition" ], "template": "___ — Manipulating a person into revealing information or taking an unsafe action", "answer": "Social engineering", "distractors": [ "High-stakes use", "De-identification", "Corroboration" ], "derivedFrom": "afe-d-manipulation", "explanation": "Compare: High-stakes use — An application where errors can materially affect safety, rights, health, livelihood, or access. De-identification — Reducing the link between data and the person it describes.", "uid": "xrhdlkxrllo" }, { "id": "czr-ai-for-everyone-afe-d-choose-tools", "shape": "cloze", "tags": [ "personal-workflow", "choose-tools", "definition" ], "template": "___ — A structured comparison of products against task, quality, risk, control, and cost criteria", "answer": "Tool evaluation", "distractors": [ "Reversible workflow", "Prompt template", "Evaluation rubric" ], "derivedFrom": "afe-d-choose-tools", "explanation": "Compare: Reversible workflow — A process whose outputs or changes can be safely reviewed and undone. Prompt template — A reusable prompt structure with named variables and consistent instructions.", "uid": "1blg3u81v2derw" }, { "id": "czr-ai-for-everyone-afe-d-evaluate", "shape": "cloze", "tags": [ "personal-workflow", "evaluate", "definition" ], "template": "___ — A set of dimensions and standards used to score outputs consistently", "answer": "Evaluation rubric", "distractors": [ "Reversible workflow", "Tool evaluation", "Prompt template" ], "derivedFrom": "afe-d-evaluate", "explanation": "Compare: Reversible workflow — A process whose outputs or changes can be safely reviewed and undone. Tool evaluation — A structured comparison of products against task, quality, risk, control, and cost criteria.", "uid": "7zd7beh0bcg6" }, { "id": "tf-t-ai-for-everyone-afe-p-name-the-task-1", "shape": "trueFalse", "tags": [ "prompt-anatomy", "name-the-task", "vocabulary" ], "statement": "\"Task verb\" means \"The action the system should perform\".", "isTrue": true, "why": "Task verb means The action the system should perform.", "derivedFrom": "afe-p-name-the-task-1", "uid": "ox4bfa1u1t8vq" }, { "id": "tf-t-ai-for-everyone-afe-p-set-constraints-1", "shape": "trueFalse", "tags": [ "prompt-anatomy", "set-constraints", "vocabulary" ], "statement": "\"Hard constraint\" means \"A boundary the output must satisfy\".", "isTrue": true, "why": "Hard constraint means A boundary the output must satisfy.", "derivedFrom": "afe-p-set-constraints-1", "uid": "1u457maweihhe" }, { "id": "tf-t-ai-for-everyone-afe-p-zero-shot-1", "shape": "trueFalse", "tags": [ "learn-by-showing", "zero-shot", "vocabulary" ], "statement": "\"Zero-shot prompt\" means \"Instructions without a worked example\".", "isTrue": true, "why": "Zero-shot prompt means Instructions without a worked example.", "derivedFrom": "afe-p-zero-shot-1", "uid": "1wjbnm8q0sqhk" }, { "id": "tf-t-ai-for-everyone-afe-p-few-shot-1", "shape": "trueFalse", "tags": [ "learn-by-showing", "few-shot", "vocabulary" ], "statement": "\"Few-shot prompt\" means \"Instructions plus several demonstrations\".", "isTrue": true, "why": "Few-shot prompt means Instructions plus several demonstrations.", "derivedFrom": "afe-p-few-shot-1", "uid": "pazvc918zl0e7" }, { "id": "tf-f-ai-for-everyone-afe-p-role-prompts-1", "shape": "trueFalse", "tags": [ "roles-reasoning", "role-prompts", "vocabulary" ], "statement": "\"Role prompt\" means \"Checks output against requirements and evidence\".", "isTrue": false, "why": "Role prompt actually means An instruction to adopt a perspective or standard.", "derivedFrom": "afe-p-role-prompts-1", "uid": "1ljis2d1w1l49b" }, { "id": "tf-t-ai-for-everyone-afe-p-decompose-1", "shape": "trueFalse", "tags": [ "roles-reasoning", "decompose", "vocabulary" ], "statement": "\"Planning stage\" means \"Chooses an approach before production\".", "isTrue": true, "why": "Planning stage means Chooses an approach before production.", "derivedFrom": "afe-p-decompose-1", "uid": "pt6pps18t0b" }, { "id": "tf-f-ai-for-everyone-afe-p-agent-loop-1", "shape": "trueFalse", "tags": [ "agents", "agent-loop", "vocabulary" ], "statement": "\"Agent observation\" means \"A change that can be safely undone\".", "isTrue": false, "why": "Agent observation actually means The result returned after an action.", "derivedFrom": "afe-p-agent-loop-1", "uid": "kjgf8r1yw1lwx" }, { "id": "tf-t-ai-for-everyone-afe-p-approval-checkpoints-1", "shape": "trueFalse", "tags": [ "agents", "approval-checkpoints", "vocabulary" ], "statement": "\"Approval gate\" means \"A required human decision before an action\".", "isTrue": true, "why": "Approval gate means A required human decision before an action.", "derivedFrom": "afe-p-approval-checkpoints-1", "uid": "1a709n8y2sqps" }, { "id": "tf-t-ai-for-everyone-afe-p-image-prompts-1", "shape": "trueFalse", "tags": [ "multimodal", "image-prompts", "vocabulary" ], "statement": "\"Image subject\" means \"The main person, object, or scene\".", "isTrue": true, "why": "Image subject means The main person, object, or scene.", "derivedFrom": "afe-p-image-prompts-1", "uid": "1awl9lv6i7055" }, { "id": "tf-f-ai-for-everyone-afe-p-draft-edit-1", "shape": "trueFalse", "tags": [ "writing-documents", "draft-edit", "vocabulary" ], "statement": "\"Developmental edit\" means \"A concise record of material revisions\".", "isTrue": false, "why": "Developmental edit actually means Changes structure, argument, or coverage.", "derivedFrom": "afe-p-draft-edit-1", "uid": "pyv1601yfb0go" }, { "id": "tf-f-ai-for-everyone-afe-p-transform-1", "shape": "trueFalse", "tags": [ "writing-documents", "transform", "vocabulary" ], "statement": "\"Semantic invariant\" means \"Recasting content into a new structure\".", "isTrue": false, "why": "Semantic invariant actually means Meaning that must remain unchanged.", "derivedFrom": "afe-p-transform-1", "uid": "1ysgep71hwy78l" }, { "id": "tf-t-ai-for-everyone-afe-p-research-plan-1", "shape": "trueFalse", "tags": [ "research-learning", "research-plan", "vocabulary" ], "statement": "\"Primary source\" means \"Original evidence or an authoritative first-party record\".", "isTrue": true, "why": "Primary source means Original evidence or an authoritative first-party record.", "derivedFrom": "afe-p-research-plan-1", "uid": "o67jot19z12tn" }, { "id": "tf-f-ai-for-everyone-afe-p-decision-support-1", "shape": "trueFalse", "tags": [ "research-learning", "decision-support", "vocabulary" ], "statement": "\"Must-have criterion\" means \"Analysis or synthesis based on other sources\".", "isTrue": false, "why": "Must-have criterion actually means A requirement an option cannot trade away.", "derivedFrom": "afe-p-decision-support-1", "uid": "h5dacuxys0g2" }, { "id": "tf-t-ai-for-everyone-afe-p-formulas-charts-1", "shape": "trueFalse", "tags": [ "data-code", "formulas-charts", "vocabulary" ], "statement": "\"Absolute reference\" means \"A spreadsheet reference fixed when copied\".", "isTrue": true, "why": "Absolute reference means A spreadsheet reference fixed when copied.", "derivedFrom": "afe-p-formulas-charts-1", "uid": "2929udn1szc7" }, { "id": "tf-f-ai-for-everyone-afe-p-test-secure-1", "shape": "trueFalse", "tags": [ "data-code", "test-secure", "vocabulary" ], "statement": "\"Sandbox\" means \"Analysis performed without running the program\".", "isTrue": false, "why": "Sandbox actually means An isolated environment that limits a program's effects.", "derivedFrom": "afe-p-test-secure-1", "uid": "14c47zlhyls9z" }, { "id": "tf-f-ai-for-everyone-afe-p-privacy-1", "shape": "trueFalse", "tags": [ "verify-protect", "privacy", "vocabulary" ], "statement": "\"Personal data\" means \"Power and accountability to make the final choice\".", "isTrue": false, "why": "Personal data actually means Information linked or linkable to a person.", "derivedFrom": "afe-p-privacy-1", "uid": "1is5co81qd178c" }, { "id": "tf-f-ai-for-everyone-afe-p-high-stakes-1", "shape": "trueFalse", "tags": [ "verify-protect", "high-stakes", "vocabulary" ], "statement": "\"Decision support\" means \"Pressure designed to prevent careful verification\".", "isTrue": false, "why": "Decision support actually means Information that helps a responsible person decide.", "derivedFrom": "afe-p-high-stakes-1", "uid": "39rj6r1lig6bx" }, { "id": "tf-t-ai-for-everyone-afe-p-templates-1", "shape": "trueFalse", "tags": [ "personal-workflow", "templates", "vocabulary" ], "statement": "\"Template variable\" means \"A field replaced for each run\".", "isTrue": true, "why": "Template variable means A field replaced for each run.", "derivedFrom": "afe-p-templates-1", "uid": "q9oaf51y7fcxn" }, { "id": "tf-t-ai-for-everyone-afe-p-handoffs-1", "shape": "trueFalse", "tags": [ "personal-workflow", "handoffs", "vocabulary" ], "statement": "\"Bounded pilot\" means \"A limited trial with success and stop criteria\".", "isTrue": true, "why": "Bounded pilot means A limited trial with success and stop criteria.", "derivedFrom": "afe-p-handoffs-1", "uid": "vj94v815ca714" }, { "id": "ai-fact-narrow-systems", "shape": "fact", "title": "Today's AI Is Bounded", "body": "Current AI systems can be broad and impressive, yet each remains a designed system with particular training, interfaces, and limits. Fluency across many tasks is not proof of human-like general intelligence.", "factVariant": "image-heavy", "imageCaption": "Impressive range is still not the same thing as unlimited intelligence.", "illustration": { "kind": "photo", "imageSearchTerm": "specialized tools workbench", "imagePrompt": "Editorial overhead photograph of a workbench holding several precise specialized tools, each suited to a different task, arranged cleanly but naturally. Warm daylight, tactile metal and wood textures, generous framing, no people, no labels, no logos, no readable text.", "alt": "A workbench with several specialized tools", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Specialized_Lululemon_car_at_Thuringen_Rundfahrt_2014.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-narrow-systems.webp" }, "tags": [ "ai-foundations", "narrow-ai", "agi" ], "uid": "1jeu2xcd1yfk" }, { "id": "ai-fact-agi-asi", "shape": "fact", "title": "AGI and Superintelligence Are Proposals", "body": "AGI usually means flexible, human-level performance across a very wide range of intellectual tasks. Superintelligence means capability far beyond humans. Neither term has one universally accepted test, and neither describes a settled present-day achievement.", "factVariant": "image-heavy", "imageCaption": "AGI and superintelligence are debated thresholds—not present-day product labels.", "illustration": { "kind": "photo", "imageSearchTerm": "foggy mountain horizon", "imagePrompt": "Cinematic editorial photograph of a distant mountain ridgeline disappearing into morning fog, with a clear foreground path and an uncertain horizon. Natural muted colors, contemplative mood, realistic landscape texture, no people, no signs, no logos, no readable text.", "alt": "A mountain path fading into a foggy horizon", "credit": "Unsplash · Sergio Vilches", "creditUrl": "https://unsplash.com/photos/silhouette-of-mountain-under-blue-sky-during-daytime-YqKS394wUc4", "url": "https://cdn.recurxive.com/packs/ai-start-here/images/ai-fact-agi-asi.webp" }, "tags": [ "ai-foundations", "agi", "superintelligence" ], "uid": "1yht8vjbr6snp" }, { "id": "ai-cmp-fluency-general", "shape": "comparison", "prompt": "Which interpretation of a chatbot's broad fluency is more careful?", "correct": { "caption": "Broad output is evidence of capability, not by itself proof of AGI.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-fluency-general-broad-output-evidence.svg" }, "incorrect": { "caption": "Broad output alone proves the system truly understands like a person.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-fluency-general-broad-output-proves.svg" }, "explanation": "Observable performance matters, but AGI claims also depend on definitions, reliability, transfer, autonomy, and the test being used.", "tags": [ "agi", "comparison", "misconception" ], "uid": "sa63q31ikd16d" }, { "id": "ai-pair-broad-fluency", "shape": "pair", "sideA": { "modality": "text", "value": "Broad fluency" }, "sideB": { "modality": "text", "value": "Evidence of capability, not proof of an agreed AGI threshold", "short": "Not proof of AGI" }, "tags": [ "agi", "vocabulary", "misconception" ], "uid": "1sey9k81gqg74" }, { "id": "ai-mcq-alphago-boundary", "shape": "mcq", "prompt": { "modality": "text", "value": "Why is a superhuman Go program still a bounded system?" }, "options": [ { "modality": "text", "value": "It uses neural networks to evaluate positions", "short": "Evaluates positions" }, { "modality": "text", "value": "It can defeat experts within the game", "short": "Beats experts at Go" }, { "modality": "text", "value": "It searches many possible move sequences", "short": "Searches move trees" }, { "modality": "text", "value": "Its skill does not transfer freely", "short": "No free skill transfer" } ], "correctIndex": 3, "explanation": "Exceptional performance in one domain does not imply flexible competence across unrelated domains.", "tags": [ "narrow-ai", "capabilities" ], "uid": "18hx8fg3zmno4" }, { "id": "ai-mcq-superintelligence-status", "shape": "mcq", "prompt": { "modality": "text", "value": "Which status best fits superintelligence today?" }, "options": [ { "modality": "text", "value": "A benchmark label for any model that beats one expert", "short": "Any expert-beating AI" }, { "modality": "text", "value": "A product term for general-purpose assistants", "short": "Assistant product term" }, { "modality": "text", "value": "A speculative future category", "short": "Speculative future" }, { "modality": "text", "value": "A regulatory category for high-risk software", "short": "High-risk software" } ], "correctIndex": 2, "explanation": "Superintelligence is a hypothetical category used in forecasting and risk debates.", "tags": [ "superintelligence", "forecasting" ], "uid": "ipn6c5r36373" }, { "id": "ai-fact-training", "shape": "fact", "title": "Training Changes the Model", "body": "During training, an optimization process repeatedly compares model outputs with a learning signal and adjusts internal weights. Training may use labeled examples, self-supervised objectives, feedback, or several stages.", "factVariant": "image-heavy", "imageCaption": "Training is the phase that adjusts the model's internal weights.", "illustration": { "kind": "photo", "imageSearchTerm": "machine learning server lab", "imagePrompt": "Documentary-style photograph inside a modern computing lab, showing engineers monitoring rows of high-performance servers during a long training run. Realistic cables, cooling hardware, subdued status lights, focused atmosphere, no visible brands, no futuristic holograms, no readable screens.", "alt": "Engineers monitoring high-performance servers in a computing lab", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Machine_Head_(band).jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-training.webp" }, "tags": [ "training", "model", "parameters" ], "uid": "jkwzb81894ksg" }, { "id": "ai-fact-inference", "shape": "fact", "title": "Inference Uses the Model", "body": "Inference is a trained model producing an output from a new input. A prompt can change what the model considers in that interaction without automatically rewriting its underlying weights.", "illustration": { "kind": "photo", "imageSearchTerm": "person typing laptop request", "imagePrompt": "Natural editorial photograph of one adult typing a request on a laptop at a quiet desk, then reviewing the result. Soft window light, realistic home-office details, screen content unreadable, calm everyday mood, no visible logos, no futuristic interface.", "alt": "A person entering a request on a laptop", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Brother_and_father_Wilfred_Marimo_in_marondera_-_march_2026.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-inference.webp" }, "tags": [ "inference", "model", "prompting" ], "uid": "1vnmwpb5fpa1p" }, { "id": "ai-cmp-train-infer", "shape": "comparison", "prompt": "Which sequence correctly connects the two phases?", "correct": { "caption": "Training adjusts weights; inference uses them to produce an output.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-train-infer-training-adjusts-weights.svg" }, "incorrect": { "caption": "Inference rebuilds the model from scratch for every prompt.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-train-infer-inference-rebuilds-model.svg" }, "explanation": "Training creates or updates learned parameters. Inference applies the resulting model.", "tags": [ "training", "inference", "comparison" ], "uid": "uwhv035vl4hl" }, { "id": "ai-pair-inference-weights", "shape": "pair", "sideA": { "modality": "text", "value": "Inference" }, "sideB": { "modality": "text", "value": "Using learned weights to answer without retraining them", "short": "Uses existing weights" }, "tags": [ "training", "inference", "vocabulary" ], "uid": "333pei1a4cn22" }, { "id": "ai-mcq-weights-change", "shape": "mcq", "prompt": { "modality": "text", "value": "In the basic two-phase model, when are learned weights adjusted?" }, "options": [ { "modality": "text", "value": "During decoding after each generated token", "short": "During token decoding" }, { "modality": "text", "value": "During training", "short": "During training" }, { "modality": "text", "value": "During inference for each prompt", "short": "Inference per prompt" }, { "modality": "text", "value": "During tokenization before inference", "short": "Before inference" } ], "correctIndex": 1, "explanation": "Optimization during training changes weights; ordinary inference reads and applies them.", "tags": [ "training", "parameters" ], "uid": "1hskrsgcfrfcg" }, { "id": "ai-mcq-query-phase", "shape": "mcq", "prompt": { "modality": "text", "value": "A user asks a deployed model to summarize a memo. What phase is this?" }, "options": [ { "modality": "text", "value": "Pretraining", "short": "Pretraining" }, { "modality": "text", "value": "Fine-tuning", "short": "Fine-tuning" }, { "modality": "text", "value": "Inference", "short": "Inference" }, { "modality": "text", "value": "Evaluation", "short": "Evaluation" } ], "correctIndex": 2, "explanation": "Generating an answer from a new request is inference.", "tags": [ "inference", "use-case" ], "uid": "rksfhy1t0wkdi" }, { "id": "ai-num-dartmouth", "shape": "numeric", "prompt": { "modality": "text", "value": "Dartmouth Summer Research Project that named the AI field" }, "value": 1956, "unit": "year", "tolerance": 0, "source": { "label": "Dartmouth — Artificial Intelligence coined at Dartmouth", "url": "https://home.dartmouth.edu/about/artificial-intelligence-ai-coined-dartmouth" }, "tags": [ "history", "dartmouth", "numeric" ], "uid": "85i41j3gma6h" }, { "id": "ai-num-deep-blue", "shape": "numeric", "prompt": { "modality": "text", "value": "IBM Deep Blue defeated chess champion Garry Kasparov" }, "value": 1997, "unit": "year", "tolerance": 0, "source": { "label": "IBM — Deep Blue", "url": "https://www.ibm.com/history/deep-blue" }, "tags": [ "history", "narrow-ai", "numeric" ], "uid": "dgac77j982l9" }, { "id": "ai-num-alexnet", "shape": "numeric", "prompt": { "modality": "text", "value": "AlexNet's landmark ImageNet competition victory" }, "value": 2012, "unit": "year", "tolerance": 0, "source": { "label": "Krizhevsky, Sutskever, and Hinton — ImageNet Classification", "url": "https://proceedings.neurips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks" }, "tags": [ "history", "deep-learning", "numeric" ], "uid": "1fabuwjwqnixl" }, { "id": "ai-num-alphago", "shape": "numeric", "prompt": { "modality": "text", "value": "AlphaGo defeated Lee Sedol in a five-game match" }, "value": 2016, "unit": "year", "tolerance": 0, "source": { "label": "Google DeepMind — AlphaGo", "url": "https://deepmind.google/research/breakthroughs/alphago/" }, "tags": [ "history", "narrow-ai", "numeric" ], "uid": "1t80ftorlc6ik" }, { "id": "ai-num-transformer", "shape": "numeric", "prompt": { "modality": "text", "value": "Publication of “Attention Is All You Need”" }, "value": 2017, "unit": "year", "tolerance": 0, "source": { "label": "Vaswani et al. — Attention Is All You Need", "url": "https://arxiv.org/abs/1706.03762" }, "tags": [ "history", "transformer", "numeric" ], "uid": "mqnqfapcn41e" }, { "id": "ai-num-chatgpt", "shape": "numeric", "prompt": { "modality": "text", "value": "Public launch of ChatGPT" }, "value": 2022, "unit": "year", "tolerance": 0, "source": { "label": "OpenAI — Introducing ChatGPT", "url": "https://openai.com/index/chatgpt/" }, "tags": [ "history", "generative-ai", "numeric" ], "uid": "scomyepyakke" }, { "id": "ai-mcq-ai-winter", "shape": "mcq", "prompt": { "modality": "text", "value": "What does an “AI winter” describe?" }, "options": [ { "modality": "text", "value": "Reduced funding and enthusiasm", "short": "Funding interest falls" }, { "modality": "text", "value": "A period of slower chip production", "short": "Slower chip production" }, { "modality": "text", "value": "A pause while new training data is collected", "short": "Pause for new data" }, { "modality": "text", "value": "A temporary decline in deployed-model accuracy", "short": "Model accuracy drops" } ], "correctIndex": 0, "explanation": "The phrase refers to downturns after expectations outran results.", "tags": [ "history", "ai-winter" ], "uid": "alyiih1xjkyoj" }, { "id": "ai-fact-model", "shape": "fact", "title": "A Model Is Learned Behavior", "body": "A machine-learning model is a mathematical system whose behavior is shaped from data. It maps inputs to outputs using learned values rather than a programmer writing every possible answer.", "illustration": { "kind": "photo", "imageSearchTerm": "machine learning model abstract", "imagePrompt": "An abstract web of glowing connected nodes and lines transforms simple input shapes into new output shapes, evoking a learned mathematical system.", "alt": "A data scientist studying patterns on a whiteboard", "depictable": false }, "tags": [ "model", "ai-foundations", "vocabulary" ], "uid": "1oxdyb41s18cn4" }, { "id": "ai-fact-data-parameters", "shape": "fact", "title": "Data and Parameters Play Different Roles", "body": "A dataset contains examples used for learning or evaluation. Parameters are adjustable internal values. Training data influences those values, but the finished model is not simply a searchable copy of its dataset.", "illustration": { "kind": "photo", "imageSearchTerm": "organized data cards desk", "imagePrompt": "Top-down editorial photograph of many varied example cards being organized beside a compact mechanical device with adjustable dials, visually suggesting examples shaping settings. Neutral desk, soft daylight, tactile paper and metal, no readable text, no logos.", "alt": "Example cards arranged beside adjustable controls", "credit": "Pexels · Laptop displaying financial chart alongside documents and credit cards on desk.", "creditUrl": "https://www.pexels.com/photo/businessman-people-laptop-office-8062355/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-data-parameters.webp" }, "tags": [ "training-data", "parameters", "model" ], "uid": "176t5fmzg192q" }, { "id": "ai-cmp-data-parameters", "shape": "comparison", "prompt": "Which distinction is correct?", "correct": { "caption": "Data supplies examples; optimization adjusts parameters.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-data-parameters-data-supplies-examples.svg" }, "incorrect": { "caption": "Parameters are the documents stored in a dataset.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-data-parameters-parameters-documents-stored.svg" }, "explanation": "Training data and learned parameters are related but not interchangeable.", "tags": [ "training-data", "parameters", "comparison" ], "uid": "bk0981mfbq8r" }, { "id": "ai-pair-parameter-count", "shape": "pair", "sideA": { "modality": "text", "value": "Parameter count" }, "sideB": { "modality": "text", "value": "A capacity measure, not a count of verified facts", "short": "Not a fact count" }, "tags": [ "parameters", "model", "misconception" ], "uid": "fo7npbrxuajd" }, { "id": "ai-mcq-dataset-role", "shape": "mcq", "prompt": { "modality": "text", "value": "What is the main role of a training dataset?" }, "options": [ { "modality": "text", "value": "Store every future answer verbatim", "short": "Stores future answers" }, { "modality": "text", "value": "Set deployment permissions", "short": "Sets permissions" }, { "modality": "text", "value": "Choose the model architecture", "short": "Chooses architecture" }, { "modality": "text", "value": "Provide learning examples", "short": "Provides examples" } ], "correctIndex": 3, "explanation": "Examples provide the signal from which the training process shapes model behavior.", "tags": [ "training-data", "training" ], "uid": "1yw42aphfvmt7" }, { "id": "ai-mcq-parameter-role", "shape": "mcq", "prompt": { "modality": "text", "value": "Which description best fits model parameters?" }, "options": [ { "modality": "text", "value": "Learned internal values", "short": "Learned model values" }, { "modality": "text", "value": "External rules attached after training", "short": "Post-training rules" }, { "modality": "text", "value": "Instructions typed by a user at inference time", "short": "User's inference text" }, { "modality": "text", "value": "Raw training examples stored verbatim", "short": "Stores raw examples" } ], "correctIndex": 0, "explanation": "Parameters are numerical values adjusted during training.", "tags": [ "parameters", "model" ], "uid": "oy0flblnepu5" }, { "id": "ai-fact-hallucination", "shape": "fact", "title": "Fluent Can Still Be False", "body": "A hallucination is generated content that is unsupported or false while sounding plausible. Examples include an invented citation, a fabricated quote, or an incorrect detail supplied without a reliable basis.", "factVariant": "image-heavy", "imageCaption": "Polished wording can hide an answer with no factual foundation.", "illustration": { "kind": "photo", "imageSearchTerm": "AI generated false information", "imagePrompt": "A page of ordinary printed text shows one sentence glowing and subtly distorted, as if a fabricated detail is hidden among truthful ones.", "alt": "A researcher checking a polished draft against source documents", "depictable": false, "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-hallucination.webp", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Edmund_Blair_Leighton_-_A_summer_shower.jpg" }, "tags": [ "hallucination", "reliability", "vocabulary" ], "uid": "201g3n1nxh851" }, { "id": "ai-fact-bias", "shape": "fact", "title": "Bias Is Often Systematic", "body": "AI bias is a pattern of unequal or distorted outcomes tied to data, design, measurement, or deployment. Unlike a one-off fabrication, bias can repeatedly disadvantage the same group or viewpoint.", "illustration": { "kind": "photo", "imageSearchTerm": "unequal outcomes tilted balance", "imagePrompt": "A balance scale tips unevenly, with one side sinking lower than the other, symbolizing a repeated pattern of unequal outcomes.", "alt": "A diverse team reviewing anonymized candidate folders", "depictable": false, "credit": "Pexels · DS stories · Pexels License", "creditUrl": "https://www.pexels.com/photo/human-shape-blocks-on-wooden-seesaw-6990404/", "subject": "Wooden toy seesaw on a white background, three yellow peg figures pressing the right end down and one red peg figure lifted on the left end", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-bias.webp" }, "tags": [ "bias", "fairness", "reliability" ], "uid": "wf3b6n1giq2lt" }, { "id": "ai-cmp-hallucination-bias", "shape": "comparison", "prompt": "Which example is best classified as a hallucination?", "correct": { "caption": "The model invents a study and a convincing-looking citation.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-hallucination-bias-model-invents-study.svg" }, "incorrect": { "caption": "A hiring system repeatedly underrates one demographic group.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-hallucination-bias-hiring-system-repeatedly.svg" }, "explanation": "Inventing a study and a citation that do not exist is hallucination — fabricated content. A hiring system that repeatedly underrates one group is bias — a systematic fairness problem, not a fabrication.", "tags": [ "hallucination", "bias", "comparison" ], "uid": "5nr98g1k6xnlo" }, { "id": "ai-pair-grounding-limits", "shape": "pair", "sideA": { "modality": "text", "value": "Grounding" }, "sideB": { "modality": "text", "value": "Relevant evidence that reduces, but cannot eliminate, hallucinations", "short": "Reduces hallucinations" }, "tags": [ "grounding", "hallucination", "vocabulary" ], "uid": "13rok9s1pthnxk" }, { "id": "ai-mcq-invented-source", "shape": "mcq", "prompt": { "modality": "text", "value": "A model supplies a plausible citation that does not exist. What happened?" }, "options": [ { "modality": "text", "value": "Hallucination", "short": "Hallucination" }, { "modality": "text", "value": "Bias", "short": "Bias" }, { "modality": "text", "value": "Prompt injection", "short": "Prompt injection" }, { "modality": "text", "value": "Data leakage", "short": "Data leakage" } ], "correctIndex": 0, "explanation": "The system generated unsupported information in a credible form.", "tags": [ "hallucination", "verification" ], "uid": "1bitbkofetbvw" }, { "id": "ai-mcq-grounding", "shape": "mcq", "prompt": { "modality": "text", "value": "What practice gives a model relevant evidence for a response?" }, "options": [ { "modality": "text", "value": "Grounding", "short": "Grounding" }, { "modality": "text", "value": "Quantization", "short": "Quantization" }, { "modality": "text", "value": "Calibration", "short": "Calibration" }, { "modality": "text", "value": "Distillation", "short": "Distillation" } ], "correctIndex": 0, "explanation": "Grounding connects generation to supplied or retrieved evidence.", "tags": [ "grounding", "reliability" ], "uid": "1kdoe6swd3tyo" }, { "id": "ai-concept-model", "shape": "concept", "conceptKind": "thing", "name": "Model", "clues": [ "A mathematical system whose behavior is shaped from examples.", "Its internal parameters are adjusted during training.", "At inference time, it maps a new input to an output." ], "tags": [ "key-concept", "model", "vocabulary" ], "uid": "10su0ez1rix0jx" }, { "id": "ai-concept-chatbot", "shape": "concept", "conceptKind": "thing", "name": "Chatbot", "clues": [ "A product interface organized around a conversation.", "It may combine a language system with retrieval, memory, and safety filters.", "It usually waits for the user's next message rather than acting through a long goal loop." ], "tags": [ "key-concept", "chatbot", "vocabulary" ], "uid": "1ilupr4178c9p8" }, { "id": "ai-concept-agent", "shape": "concept", "conceptKind": "thing", "name": "AI agent", "clues": [ "A goal-directed system that can operate across several steps.", "It may keep state, select tools, inspect results, and change its next action.", "Its loop commonly follows observe, plan, act, check, and stop or continue." ], "tags": [ "key-concept", "agent", "tool-use" ], "uid": "kr1z361cx426" }, { "id": "ai-concept-hallucination", "shape": "concept", "conceptKind": "thing", "name": "Hallucination", "clues": [ "A reliability failure whose surface can look polished and plausible.", "It occurs when generated content is unsupported or false.", "An invented study or nonexistent citation is a classic example." ], "tags": [ "key-concept", "hallucination", "reliability" ], "uid": "dp27g518tnbnr" }, { "id": "ai-concept-grounding", "shape": "concept", "conceptKind": "thing", "name": "Grounding", "clues": [ "A reliability technique that gives generation an evidence base.", "The evidence may come from supplied documents, retrieval, or a trusted database.", "It reduces unsupported answers but still requires checking." ], "tags": [ "key-concept", "grounding", "reliability" ], "uid": "1ioq37fcmcqx1" }, { "id": "ai-concept-anthropomorphism", "shape": "concept", "conceptKind": "thing", "name": "Anthropomorphism", "clues": [ "A reasoning trap that can be triggered by a humanlike interface.", "It treats convenient behavioral language as evidence of an inner experience.", "It means attributing human traits or motives to a nonhuman system." ], "tags": [ "key-concept", "anthropomorphism", "misconception" ], "uid": "w547ejdetrs5" }, { "id": "ai-mcq-agent-addition", "shape": "mcq", "prompt": { "modality": "text", "value": "Which feature most clearly moves a chatbot toward an agent?" }, "options": [ { "modality": "text", "value": "A more conversational tone", "short": "Conversational tone" }, { "modality": "text", "value": "More response-formatting templates", "short": "Formatting templates" }, { "modality": "text", "value": "A longer context window", "short": "Longer context" }, { "modality": "text", "value": "Multi-step tool use", "short": "Multi-step tool use" } ], "correctIndex": 3, "explanation": "Planning and executing actions through tools is a central agentic capability.", "tags": [ "agent", "tool-use" ], "uid": "dhyzyj75q5wh" }, { "id": "ai-fact-strengths", "shape": "fact", "title": "AI Excels at Pattern-Rich Draft Work", "body": "Current generative systems are often useful for drafting, summarizing, extracting, classifying, translating, coding assistance, and generating alternatives—especially when the task is well specified and easy to review.", "factVariant": "image-heavy", "imageCaption": "The sweet spot: structured, reversible work that a person can review.", "illustration": { "kind": "photo", "imageSearchTerm": "writer reviewing drafts laptop", "imagePrompt": "Natural editorial photograph of a professional comparing several draft pages beside a laptop, sorting and revising them with a pen. Bright but realistic workspace, relaxed concentration, pages contain no readable text, no visible brands, no futuristic effects.", "alt": "A professional reviewing several drafts beside a laptop", "credit": "Unsplash · Vitaly Gariev · Unsplash License", "creditUrl": "https://unsplash.com/photos/woman-working-with-documents-at-office-desk-Vvn0OD0IxBM", "subject": "Woman in glasses at an office desk holding a clipboard and lifting a printed page with a chart on it, open laptop in front of her, notebook with handwritten notes and pen on the desk", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-strengths.webp" }, "tags": [ "capabilities", "use-case", "reliability" ], "uid": "11g60ugaho3fk" }, { "id": "ai-fact-reliability-gap", "shape": "fact", "title": "Confidence Is a Style, Not a Score", "body": "A model can express a wrong answer in polished, decisive language. Reliability comes from evidence, tests, monitoring, and human judgment—not from tone.", "factVariant": "image-heavy", "imageCaption": "Confidence in the prose is not confidence in the facts.", "illustration": { "kind": "photo", "imageSearchTerm": "editor verifying report sources", "imagePrompt": "Editorial photograph of an experienced editor carefully verifying a polished report against multiple source documents, calculator, and notebook. Focused expression, natural office light, tactile paper detail, all page text unreadable, no logos, no staged futuristic technology.", "alt": "An editor verifying a polished report against source material", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Joe%27s_Own_Editor_screenshot.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-reliability-gap.webp" }, "tags": [ "reliability", "verification", "hallucination" ], "uid": "uu5dfo134orw4" }, { "id": "ai-cmp-task-fit", "shape": "comparison", "prompt": "Which is the safer first use of generative AI?", "correct": { "caption": "Draft three reversible headlines for a person to review.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-task-fit-draft-three-reversible.svg" }, "incorrect": { "caption": "Make an unsupervised medical diagnosis from incomplete notes.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-task-fit-make-unsupervised-medical.svg" }, "explanation": "Reversible, reviewable work tolerates error better than high-stakes decisions.", "tags": [ "capabilities", "high-stakes", "comparison" ], "uid": "1v8x8c9do3ld7" }, { "id": "ai-pair-confident-tone", "shape": "pair", "sideA": { "modality": "text", "value": "Confident tone" }, "sideB": { "modality": "text", "value": "A writing style, not evidence that an answer is correct", "short": "Style, not proof" }, "tags": [ "reliability", "verification", "misconception" ], "uid": "14x8wunrpjxp1" }, { "id": "ai-mcq-first-use", "shape": "mcq", "prompt": { "modality": "text", "value": "Which task is the best beginner pilot?" }, "options": [ { "modality": "text", "value": "An automatically approved customer refund", "short": "Auto-approved refund" }, { "modality": "text", "value": "A final résumé-screening rejection", "short": "Final résumé rejection" }, { "modality": "text", "value": "An unsupervised compliance filing", "short": "Unsupervised filing" }, { "modality": "text", "value": "A reviewable first draft", "short": "Reviewable draft" } ], "correctIndex": 3, "explanation": "A reversible draft creates value while keeping the consequence of an error low.", "tags": [ "use-case", "human-in-the-loop" ], "uid": "1hzjx2rxy0gy9" }, { "id": "ai-mcq-reliability-signal", "shape": "mcq", "prompt": { "modality": "text", "value": "What most improves trust in an important output?" }, "options": [ { "modality": "text", "value": "Requesting a more detailed explanation", "short": "Request more detail" }, { "modality": "text", "value": "Raising the model's self-reported confidence", "short": "Higher self-confidence" }, { "modality": "text", "value": "Independent verification", "short": "Independent check" }, { "modality": "text", "value": "Repeating the same model until answers agree", "short": "Repeat until agreement" } ], "correctIndex": 2, "explanation": "Checks against primary evidence or tested data address correctness directly.", "tags": [ "verification", "reliability" ], "uid": "1crzodb1b2ba81" }, { "id": "ai-fact-energy", "shape": "fact", "title": "AI Runs on Physical Infrastructure", "body": "AI computation uses servers, networking, cooling, electricity, and buildings. The IEA projects total data-center electricity demand—not AI alone—to reach about 945 TWh in 2030, with AI the largest driver of growth.", "factVariant": "image-heavy", "imageCaption": "Every digital answer depends on servers, cooling, electricity, and buildings.", "illustration": { "kind": "photo", "imageSearchTerm": "data center technician servers", "imagePrompt": "Wide documentary photograph of a technician walking between long rows of modern server racks in a real data center. Visible cabling, ventilation, cooling infrastructure, practical overhead lighting, accurate industrial scale, no brand logos, no neon science-fiction styling, no readable screens.", "alt": "A technician walking between server racks in a data center", "credit": "Pexels · A woman using a laptop navigating a contemporary data center with mirrored servers.", "creditUrl": "https://www.pexels.com/photo/engineer-holding-laptop-1181316/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-energy.webp" }, "source": { "label": "IEA — Energy and AI (2025)", "url": "https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai" }, "tags": [ "energy", "data-centers", "environment" ], "uid": "pflq5w1ncam14" }, { "id": "ai-fact-water-context", "shape": "fact", "title": "Water Use Depends on Place and Time", "body": "Data centers may use water directly for cooling and indirectly through electricity generation and chip manufacturing. The impact varies with cooling design, climate, grid mix, and local water stress, so one global per-prompt number can mislead.", "illustration": { "kind": "photo", "imageSearchTerm": "data center cooling towers", "imagePrompt": "A large data center facility has industrial cooling towers and pipes venting steam, showing the water infrastructure that cools the servers.", "alt": "Cooling pipes and water infrastructure at an industrial facility", "depictable": true, "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Bildschirmfoto_2021-02-14_um_16.32.06.png", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-water-context.webp" }, "tags": [ "water", "data-centers", "environment" ], "uid": "196ata11sbhlov" }, { "id": "ai-cmp-energy-claim", "shape": "comparison", "prompt": "Which claim handles the IEA projection accurately?", "correct": { "caption": "945 TWh is projected for all data centers; AI drives much of the growth.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-energy-claim-945-twh-projected.svg" }, "incorrect": { "caption": "Every single one of the 945 TWh is assigned exclusively to AI models.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-energy-claim-every-945-twh.svg" }, "explanation": "AI-optimized servers are a major driver, but the projection covers the whole data-center sector.", "tags": [ "energy", "data-centers", "comparison" ], "uid": "12xqabk3qsuj8" }, { "id": "ai-pair-physical-cost", "shape": "pair", "sideA": { "modality": "text", "value": "AI output" }, "sideB": { "modality": "text", "value": "A digital result produced by physical computing infrastructure", "short": "Uses physical systems" }, "tags": [ "energy", "data-centers", "environment" ], "uid": "1lejn291jme7hn" }, { "id": "ai-mcq-footprint-context", "shape": "mcq", "prompt": { "modality": "text", "value": "Why can one universal water-per-prompt number mislead?" }, "options": [ { "modality": "text", "value": "Indirect electricity-related water is always excluded", "short": "Excludes grid water" }, { "modality": "text", "value": "Every facility uses the same cooling design", "short": "Uniform cooling" }, { "modality": "text", "value": "Systems and locations differ", "short": "Systems, places vary" }, { "modality": "text", "value": "Token count is the only variable that matters", "short": "Token count alone" } ], "correctIndex": 2, "explanation": "Model, hardware, cooling, weather, grid, and accounting boundaries all affect an estimate.", "tags": [ "water", "environment" ], "uid": "1whzi7hg0pwjn" }, { "id": "ai-mcq-iea-scope", "shape": "mcq", "prompt": { "modality": "text", "value": "What sector does the IEA's 945 TWh projection cover?" }, "options": [ { "modality": "text", "value": "All data centers", "short": "All data centers" }, { "modality": "text", "value": "Cryptocurrency mining centers only", "short": "Crypto centers only" }, { "modality": "text", "value": "Data centers in advanced economies only", "short": "Wealthy economies only" }, { "modality": "text", "value": "AI-optimized data centers only", "short": "AI centers only" } ], "correctIndex": 0, "explanation": "The figure is a sector-wide data-center projection for 2030.", "source": { "label": "IEA — Energy and AI (2025)", "url": "https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai" }, "tags": [ "energy", "data-centers" ], "uid": "6pkthgdbt9vk" }, { "id": "ai-fact-eu-act", "shape": "fact", "title": "The EU Uses a Binding Risk Framework", "body": "The EU AI Act assigns obligations by risk and use. As of July 27, 2026, its Omnibus extends key high-risk-system dates while transparency duties for chatbots and some generated content begin August 2, 2026.", "factVariant": "image-heavy", "imageCaption": "The EU AI Act organizes binding duties around risk and use.", "illustration": { "kind": "photo", "imageSearchTerm": "European Union flags parliament", "imagePrompt": "Editorial photograph of European Union flags outside a modern European institutional building, photographed in clear natural daylight with architectural depth. Accurate blue-and-gold flags, restrained civic mood, no political campaign signs, no company logos, no readable text.", "alt": "European Union flags outside an institutional building", "credit": "Pexels · European Union flags waving outside the European Parliament in Strasbourg.", "creditUrl": "https://www.pexels.com/photo/european-union-flags-at-strasbourg-parliament-39049421/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-eu-act.webp" }, "source": { "label": "European Commission — AI Act framework", "url": "https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai" }, "tags": [ "regulation", "eu-ai-act", "risk" ], "uid": "1dgg63a1g21ju6" }, { "id": "ai-fact-us-policy", "shape": "fact", "title": "The U.S. Uses a More Distributed Approach", "body": "The July 2025 U.S. AI Action Plan emphasizes innovation, infrastructure, national security, and adoption. Federal agencies, sector laws, procurement rules, courts, and state laws still create additional obligations; there is no single EU-style federal AI Act.", "illustration": { "kind": "photo", "imageSearchTerm": "United States Capitol technology", "imagePrompt": "Editorial photograph of the United States Capitol viewed from a contemporary government-office setting with subtle technology context such as an open laptop in the foreground. Natural daylight, neutral civic tone, no partisan signs, no logos, screen unreadable.", "alt": "The United States Capitol with a laptop in the foreground", "credit": "Pexels · A smartphone captures the Capitol Building at twilight, highlighting iconic architecture and technology.", "creditUrl": "https://www.pexels.com/photo/person-holding-iphone-413878/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-us-policy.webp" }, "source": { "label": "White House — America's AI Action Plan", "url": "https://www.whitehouse.gov/srv/htdocs/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf" }, "tags": [ "regulation", "us-policy", "governance" ], "uid": "104qk4aomx28a" }, { "id": "ai-cmp-eu-us", "shape": "comparison", "prompt": "Which high-level comparison is more accurate in July 2026?", "correct": { "caption": "EU: one cross-sector risk law. U.S.: sectoral, agency, court, and state layers.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-eu-us-eu-cross-sector.svg" }, "incorrect": { "caption": "The EU and U.S. both rely on the exact same single federal statute and deadlines.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-eu-us-eu-u-s.svg" }, "explanation": "Both govern AI, but their legal structures and policy emphasis differ.", "tags": [ "regulation", "eu-ai-act", "us-policy", "comparison" ], "uid": "1s90zjp1h1gupf" }, { "id": "ai-pair-jurisdiction", "shape": "pair", "sideA": { "modality": "text", "value": "AI regulation" }, "sideB": { "modality": "text", "value": "Rules that vary by jurisdiction, sector, and use", "short": "Varies by jurisdiction" }, "tags": [ "regulation", "jurisdiction", "governance" ], "uid": "1l37dwr13begat" }, { "id": "ai-mcq-eu-organizing-idea", "shape": "mcq", "prompt": { "modality": "text", "value": "What organizes the EU AI Act's obligations?" }, "options": [ { "modality": "text", "value": "Model parameter count alone", "short": "Parameter count alone" }, { "modality": "text", "value": "Risk and use category", "short": "Risk and use category" }, { "modality": "text", "value": "Company revenue and employee count", "short": "Revenue and headcount" }, { "modality": "text", "value": "The developer's headquarters country", "short": "Developer home country" } ], "correctIndex": 1, "explanation": "The Act distinguishes prohibited, high-risk, transparency, and minimal-risk uses.", "source": { "label": "European Commission — AI Act framework", "url": "https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai" }, "tags": [ "regulation", "eu-ai-act" ], "uid": "1l5ry2j3byao5" }, { "id": "ai-def-ai-governance", "shape": "definition", "term": { "modality": "text", "value": "AI governance" }, "definition": { "modality": "text", "value": "The rules, roles, controls, and oversight used to direct how AI is developed and used.", "short": "AI rules and oversight" }, "curatedDistractors": [ "Inference", "Tokenization", "Benchmarking" ], "tags": [ "regulation", "governance", "vocabulary" ], "uid": "9favuvu5c391" }, { "id": "ai-fact-jobs-tasks", "shape": "fact", "title": "Tasks Change Before Whole Jobs Disappear", "body": "Jobs bundle many tasks, relationships, responsibilities, and physical constraints. AI may automate some tasks, complement others, and create new work, so “exposed” does not mean “eliminated.”", "factVariant": "image-heavy", "imageCaption": "Jobs are bundles of tasks; technology can change the bundle without erasing it.", "illustration": { "kind": "photo", "imageSearchTerm": "electrician tablet troubleshooting work", "imagePrompt": "Documentary photograph of a skilled electrician using a tablet while inspecting real electrical equipment, combining digital assistance with hands-on judgment. Natural worksite light, authentic tools and protective clothing, no visible brands, no staged holograms, screen unreadable.", "alt": "An electrician using a tablet while inspecting equipment", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Electrician_repairing_a_fun_4.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-jobs-tasks.webp" }, "tags": [ "jobs", "economics", "automation" ], "uid": "1112vwv1smq4vl" }, { "id": "ai-fact-job-estimates", "shape": "fact", "title": "Big Numbers Often Measure Different Things", "body": "The IMF estimates about 40% of global employment is exposed to AI, including work AI may complement. Daron Acemoglu models a much smaller near-term productivity effect. These claims differ because they ask different questions.", "illustration": { "kind": "photo", "imageSearchTerm": "economist comparing charts", "imagePrompt": "Editorial photograph of two economists comparing differently structured charts and research notes at a table, discussing assumptions rather than arguing. Calm academic setting, soft daylight, charts visible as shapes but numbers unreadable, no logos, no staged technology.", "alt": "Two economists comparing charts and research notes", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Noah_Yosif_Headshot.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-job-estimates.webp" }, "source": { "label": "IMF — AI and the Future of Work", "url": "https://www.elibrary.imf.org/view/journals/006/2024/001/article-A001-en.xml" }, "tags": [ "jobs", "economics", "forecasting" ], "uid": "75dyqn1q3xu2x" }, { "id": "ai-cmp-exposure-displacement", "shape": "comparison", "prompt": "Which reading of an “AI-exposed job” is correct?", "correct": { "caption": "Some important tasks could change, whether automated or assisted.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-exposure-displacement-some-important-tasks.svg" }, "incorrect": { "caption": "The entire occupation is certain to vanish within just a few short years.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-exposure-displacement-entire-occupation-certain.svg" }, "explanation": "Exposure measures overlap with AI capabilities, not a guaranteed employment outcome.", "tags": [ "jobs", "automation", "comparison" ], "uid": "1v0opf01jyg1z0" }, { "id": "ai-pair-job-exposure", "shape": "pair", "sideA": { "modality": "text", "value": "Job exposure" }, "sideB": { "modality": "text", "value": "Tasks may change; it does not guarantee job elimination", "short": "Tasks may change" }, "tags": [ "jobs", "automation", "economics" ], "uid": "z1x3d21r3zsfy" }, { "id": "ai-mcq-job-unit", "shape": "mcq", "prompt": { "modality": "text", "value": "What is usually the better first unit for analyzing automation?" }, "options": [ { "modality": "text", "value": "An industry's annual revenue", "short": "Industry revenue" }, { "modality": "text", "value": "A firm's total headcount", "short": "Firm headcount" }, { "modality": "text", "value": "An occupation treated as indivisible", "short": "Whole occupation" }, { "modality": "text", "value": "A task within a job", "short": "A task within a job" } ], "correctIndex": 3, "explanation": "Occupations combine tasks with very different potential for assistance or substitution.", "tags": [ "jobs", "automation" ], "uid": "rd014e13bmfpi" }, { "id": "ai-mcq-estimate-conflict", "shape": "mcq", "prompt": { "modality": "text", "value": "Why can two credible job-impact estimates differ sharply?" }, "options": [ { "modality": "text", "value": "One report includes more industry case studies", "short": "More industry cases" }, { "modality": "text", "value": "Definitions and assumptions differ", "short": "Definitions differ" }, { "modality": "text", "value": "One report uses a point estimate instead of a range", "short": "Point instead of range" }, { "modality": "text", "value": "One team publishes a longer methodology appendix", "short": "Longer methods section" } ], "correctIndex": 1, "explanation": "Time horizon, capability assumptions, exposure definitions, adoption, and task coverage can change the result.", "tags": [ "jobs", "forecasting" ], "uid": "1aq3rhr1vovloh" }, { "id": "ai-fact-xrisk-concern", "shape": "fact", "title": "The Concerned Case Is Forward-Looking", "body": "Researchers worried about catastrophic AI risk argue that future systems could become more capable, autonomous, and difficult to control. Because the possible harm is enormous, they favor testing, monitoring, access controls, and international coordination before evidence is conclusive.", "factVariant": "image-heavy", "imageCaption": "For high-consequence risks, uncertainty can be a reason to test before deployment.", "illustration": { "kind": "photo", "imageSearchTerm": "safety engineers control room", "imagePrompt": "Serious editorial photograph of a multidisciplinary safety team studying a complex system in a real control room before activation. People reviewing checklists and monitors, cautious professional mood, natural human expressions, screens unreadable, no robots, no logos, no science-fiction holograms.", "alt": "A safety team reviewing a complex system before activation", "credit": "Pexels · Engineers wearing safety gear operate a control panel in an Indonesian facility.", "creditUrl": "https://www.pexels.com/photo/employees-operating-industrial-machinery-10871565/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-xrisk-concern.webp" }, "tags": [ "x-risk", "safety", "governance" ], "uid": "1hg66lr1d8zh4h" }, { "id": "ai-fact-xrisk-skeptic", "shape": "fact", "title": "The Skeptical Case Starts with Present Evidence", "body": "Skeptics argue that current systems remain brittle, that forecasts extrapolate too far, and that attention to speculative catastrophe can crowd out measured harms such as fraud, bias, surveillance, labor disruption, and concentrated power.", "illustration": { "kind": "photo", "imageSearchTerm": "researcher examining evidence documents", "imagePrompt": "Editorial photograph of a skeptical researcher examining concrete evidence, incident reports, and measured results at a desk. Careful analytical posture, grounded ordinary office, realistic paper and laptop texture, text unreadable, no logos, no dramatic science-fiction imagery.", "alt": "A researcher examining incident reports and measured evidence", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:ThomasLoerting.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-xrisk-skeptic.webp" }, "tags": [ "x-risk", "safety", "present-harms" ], "uid": "8edzsynjl66i" }, { "id": "ai-cmp-xrisk-reasons", "shape": "comparison", "prompt": "Which statement fairly represents the disagreement?", "correct": { "caption": "The camps weigh future capability, uncertainty, and opportunity cost differently.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-xrisk-reasons-camps-weigh-future.svg" }, "incorrect": { "caption": "One camp genuinely cares about safety, while the other simply does not care.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-xrisk-reasons-camp-cares-about.svg" }, "explanation": "Both can care about harm while disagreeing about probability, tractability, and priorities.", "tags": [ "x-risk", "comparison", "disagreement" ], "uid": "6v6tgd1jicgvr" }, { "id": "ai-pair-xrisk-forecast", "shape": "pair", "sideA": { "modality": "text", "value": "AI x-risk forecast" }, "sideB": { "modality": "text", "value": "A contested probability judgment, not a scientific consensus", "short": "Contested probability" }, "tags": [ "x-risk", "forecasting", "disagreement" ], "uid": "6o6mqnrcazst" }, { "id": "ai-mcq-xrisk-root", "shape": "mcq", "prompt": { "modality": "text", "value": "What drives much of the catastrophic-risk disagreement?" }, "options": [ { "modality": "text", "value": "Judgments about current chatbot interfaces", "short": "Chatbot interface" }, { "modality": "text", "value": "Preferences about open-source licensing", "short": "Open-source stance" }, { "modality": "text", "value": "Views of present data-center energy use", "short": "Current energy views" }, { "modality": "text", "value": "Views of future capability growth", "short": "Future capability" } ], "correctIndex": 3, "explanation": "Forecasts turn on whether progress will stay gradual or produce systems with much greater autonomy and power.", "tags": [ "x-risk", "forecasting" ], "uid": "1sstkyv191h95" }, { "id": "ai-mcq-xpt-gap", "shape": "mcq", "prompt": { "modality": "text", "value": "In the XPT, which group gave the lower median AI-extinction forecast?" }, "options": [ { "modality": "text", "value": "General x-risk experts", "short": "General x-risk experts" }, { "modality": "text", "value": "AI domain experts", "short": "AI domain experts" }, { "modality": "text", "value": "Non-AI domain experts", "short": "Non-AI experts" }, { "modality": "text", "value": "Superforecasters", "short": "Superforecasters" } ], "correctIndex": 3, "explanation": "The 2100 medians were about 0.38% for superforecasters and 3% for AI domain experts.", "source": { "label": "Forecasting Research Institute — XPT report", "url": "https://forecastingresearch.org/s/XPT.pdf" }, "tags": [ "x-risk", "forecasting", "numeric" ], "uid": "17g1xuo1aaemms" }, { "id": "ai-fact-copyright-question", "shape": "fact", "title": "Training Raises Several Copyright Questions", "body": "In the United States, disputes can involve copying during data collection, acquiring works from unlawful sources, using them in training, memorized material, and competing outputs. One answer does not automatically settle the others.", "factVariant": "image-heavy", "imageCaption": "Acquisition, training, memorization, and outputs raise different legal questions.", "illustration": { "kind": "photo", "imageSearchTerm": "law books digital archive", "imagePrompt": "Editorial still life combining shelves of books, archival document boxes, and a laptop used for digital research on a law-library table. Thoughtful legal atmosphere, realistic materials, screen and book text unreadable, no gavels, no company logos, no futuristic effects.", "alt": "Books, archive boxes, and a laptop on a law-library table", "credit": "Pexels · Eric Lozaga · Pexels License", "creditUrl": "https://www.pexels.com/photo/modern-library-archive-storage-room-with-compact-shelving-31139000/", "subject": "a long corridor of beige mobile compact-shelving units with black crank handles, bay signs 'UNLOCKED COMPACT STORAGE 105/129', and shelves of tan-and-black bound law reporters visible on the right; no people", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-copyright-question.webp" }, "tags": [ "copyright", "training-data", "law" ], "uid": "1eepw8j1tzdt6h" }, { "id": "ai-fact-copyright-split", "shape": "fact", "title": "Facts and Markets Matter to Fair Use", "body": "U.S. courts have reached context-specific results. A 2025 ruling rejected fair use where Ross copied Westlaw headnotes for a competing legal tool; another found Anthropic's model training fair use but treated its pirated book library separately.", "illustration": { "kind": "photo", "imageSearchTerm": "judge legal documents technology", "imagePrompt": "Neutral editorial photograph of legal professionals reviewing thick case files beside a laptop in a courthouse library. Emphasis on fact-specific analysis and documents, natural indoor light, no visible case names, no readable text, no logos, no theatrical gavel.", "alt": "Legal professionals reviewing case files beside a laptop", "credit": "Pexels · Female judge in a courtroom setting, focusing on legal documents with a gavel.", "creditUrl": "https://www.pexels.com/photo/judge-in-courtroom-making-a-ruling-34817076/", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-copyright-split.webp" }, "source": { "label": "The Copyright Society — Training on Trial (2026)", "url": "https://copyrightsociety.org/wp-content/uploads/2026/05/73-J.-Copyright-Socy-261-2026.pdf" }, "tags": [ "copyright", "fair-use", "law" ], "uid": "1ofgp899zwvtn" }, { "id": "ai-cmp-copyright-claim", "shape": "comparison", "prompt": "Which July 2026 summary is legally careful?", "correct": { "caption": "U.S. outcomes remain fact-specific; acquisition, training, and outputs can differ.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-copyright-claim-u-s-outcomes.svg" }, "incorrect": { "caption": "Every unlicensed training use is now always fully legal, or always entirely illegal.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-copyright-claim-every-unlicensed-training.svg" }, "explanation": "District-court rulings and the Copyright Office show a developing, context-dependent area.", "tags": [ "copyright", "fair-use", "comparison" ], "uid": "18oy27r1jgiwp5" }, { "id": "ai-pair-acquisition-use", "shape": "pair", "sideA": { "modality": "text", "value": "Acquisition vs. use" }, "sideB": { "modality": "text", "value": "Separate legal questions even when a later use is fair", "short": "Separate legal issues" }, "tags": [ "copyright", "fair-use", "law" ], "uid": "11uxurb1xpga1x" }, { "id": "ai-mcq-copyright-status", "shape": "mcq", "prompt": { "modality": "text", "value": "What is the safest general description of U.S. training law in July 2026?" }, "options": [ { "modality": "text", "value": "Settled nationwide in favor of every unlicensed use", "short": "All uses held legal" }, { "modality": "text", "value": "Governed by one rule that ignores the training corpus", "short": "One corpus-blind rule" }, { "modality": "text", "value": "Settled nationwide against every unlicensed use", "short": "All uses held illegal" }, { "modality": "text", "value": "Developing and fact-specific", "short": "Still fact-specific" } ], "correctIndex": 3, "explanation": "Cases turn on purpose, source, market effect, output behavior, and other facts; many suits remain active.", "tags": [ "copyright", "law" ], "uid": "igbotm17ioute" }, { "id": "ai-def-fair-use", "shape": "definition", "term": { "modality": "text", "value": "Fair use" }, "definition": { "modality": "text", "value": "A U.S. doctrine that evaluates certain unlicensed uses through four statutory factors and case-specific facts.", "short": "Four-factor exception" }, "curatedDistractors": [ "Public domain", "Patent", "Trade secret" ], "tags": [ "copyright", "fair-use", "vocabulary" ], "uid": "c4yj91vfq4q7" }, { "id": "ai-fact-agi-definition", "shape": "fact", "title": "The Finish Line Moves with the Definition", "body": "Forecasts may target AGI, high-level machine intelligence, all economically valuable work, or machines outperforming humans at every task. These thresholds are not interchangeable.", "illustration": { "kind": "photo", "imageSearchTerm": "multiple finish lines track", "imagePrompt": "Conceptual editorial photograph of a running track with several finish ribbons set at different distances, showing that different definitions create different thresholds. Clean outdoor light, no athletes, no readable signs, no logos, realistic photographic perspective.", "alt": "A running track with several finish lines at different distances", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Cuerpos.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-agi-definition.webp" }, "tags": [ "agi", "forecasting", "definitions" ], "uid": "u4v7571t97txl" }, { "id": "ai-fact-agi-survey", "shape": "fact", "title": "A Survey Is a Distribution, Not a Date", "body": "A survey of 2,778 AI researchers produced an aggregate 50% date of 2047 for unaided machines outperforming humans at every possible task, while individual answers varied widely. It is a conditional forecast, not a schedule.", "factVariant": "image-heavy", "imageCaption": "One median date can hide a wide distribution of expert uncertainty.", "illustration": { "kind": "photo", "imageSearchTerm": "researchers probability forecast charts", "imagePrompt": "Editorial photograph of researchers studying a wall of overlapping probability curves and forecast ranges, emphasizing a wide distribution rather than one date. Calm academic workspace, chart labels unreadable, natural light, no logos, no futuristic interface.", "alt": "Researchers examining a range of probability forecasts", "credit": "Wikimedia Commons · see source", "creditUrl": "https://commons.wikimedia.org/wiki/File:Ahmad_Khalifah.jpg", "url": "https://cdn.recurxive.com/packs/ai-for-everyone/images/ai-fact-agi-survey.webp" }, "source": { "label": "Grace et al. — Thousands of AI Authors", "url": "https://arxiv.org/abs/2401.02843" }, "tags": [ "agi", "forecasting", "survey" ], "uid": "qd9796hmhtyu" }, { "id": "ai-cmp-agi-date", "shape": "comparison", "prompt": "Which interpretation of the 2047 survey result is sound?", "correct": { "caption": "It is an aggregate 50% survey forecast under one specific stated threshold.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-agi-date-aggregate-50-forecast.svg" }, "incorrect": { "caption": "Researchers definitively proved AGI will arrive exactly during 2047.", "imageUrl": "/packs/ai-for-everyone/diagrams/ai-cmp-agi-date-researchers-proved-agi.svg" }, "explanation": "A probabilistic survey summarizes beliefs and uncertainty; it does not establish an event date.", "tags": [ "agi", "forecasting", "comparison" ], "uid": "z7ww2odewows" }, { "id": "ai-tf-agi-countdown", "shape": "trueFalse", "statement": "An expert median date should be read as a guaranteed countdown to AGI.", "isTrue": false, "why": "Forecasts depend on definitions, methods, assumptions, and future evidence, and can shift substantially.", "tags": [ "agi", "forecasting", "misconception" ], "uid": "wav77l4e281r" }, { "id": "ai-mcq-timeline-variable", "shape": "mcq", "prompt": { "modality": "text", "value": "What should be checked before comparing two AGI timelines?" }, "options": [ { "modality": "text", "value": "Their target definitions", "short": "Target definitions" }, { "modality": "text", "value": "Their publication locations", "short": "Publication locations" }, { "modality": "text", "value": "Their source-code licenses", "short": "Source-code licenses" }, { "modality": "text", "value": "Their preferred hardware vendors", "short": "Hardware vendors" } ], "correctIndex": 0, "explanation": "Different thresholds can yield different dates even when forecasters share similar beliefs.", "tags": [ "agi", "forecasting" ], "uid": "ti2tlvuvgy3d" }, { "id": "ai-cloze-forecast", "shape": "cloze", "template": "A ___ expresses a probability judgment based on present evidence and assumptions.", "answer": "forecast", "distractors": [ "model", "dataset", "regulation" ], "explanation": "A forecast is conditional and revisable; it is not a guaranteed event date.", "tags": [ "forecasting", "vocabulary" ], "uid": "8u42f12hx7cz" }, { "id": "ai-pair-ai-winter", "shape": "pair", "sideA": { "modality": "text", "value": "AI winter" }, "sideB": { "modality": "text", "value": "A period of reduced AI funding and enthusiasm", "short": "AI funding slows" }, "tags": [ "history", "ai-winter", "vocabulary" ], "uid": "6qz2212327mi" }, { "id": "ai-pair-agent-workflow", "shape": "pair", "sideA": { "modality": "text", "value": "AI agent" }, "sideB": { "modality": "text", "value": "A system pursuing a goal across steps, tools, or actions", "short": "Acts across steps" }, "tags": [ "agent", "tool-use", "vocabulary" ], "uid": "1o0x10y1na6yyu" }, { "id": "ai-pair-median-forecast", "shape": "pair", "sideA": { "modality": "text", "value": "Median forecast" }, "sideB": { "modality": "text", "value": "The midpoint of individual probability judgments, not a deadline", "short": "Midpoint, not deadline" }, "tags": [ "forecasting", "survey", "vocabulary" ], "uid": "i0y5an1kc2k5t" }, { "id": "ai-def-agi", "shape": "definition", "term": { "modality": "text", "value": "AGI" }, "definition": { "modality": "text", "value": "A proposed threshold for broad, flexible performance across many intellectual tasks; no universal test exists.", "short": "Broad capability" }, "curatedDistractors": [ "Narrow AI", "Superintelligence", "Automation" ], "tags": [ "agi", "definitions", "vocabulary" ], "uid": "1xkbazt13moa4f" }, { "id": "ai-def-inference", "shape": "definition", "term": { "modality": "text", "value": "Inference" }, "definition": { "modality": "text", "value": "Using a trained model's learned values to produce an output for a request.", "short": "Use model to answer" }, "curatedDistractors": [ "Training", "Fine-tuning", "Data labeling" ], "tags": [ "inference", "training", "vocabulary" ], "uid": "1iijmuy1jd3hha" }, { "id": "ai-def-parameters", "shape": "definition", "term": { "modality": "text", "value": "Parameters" }, "definition": { "modality": "text", "value": "Learned internal numerical values adjusted during training to shape model behavior.", "short": "Learned values" }, "curatedDistractors": [ "Training data", "Prompts", "Tokens" ], "tags": [ "parameters", "model", "vocabulary" ], "uid": "auk4nb1jptnyl" }, { "id": "ai-def-hallucination", "shape": "definition", "term": { "modality": "text", "value": "Hallucination" }, "definition": { "modality": "text", "value": "A plausible-seeming model output that is unsupported, incorrect, or invented.", "short": "Unsupported output" }, "curatedDistractors": [ "Bias", "Data leakage", "Prompt injection" ], "tags": [ "hallucination", "reliability", "vocabulary" ], "uid": "77nlky1qeafvi" }, { "id": "ai-def-agent", "shape": "definition", "term": { "modality": "text", "value": "AI agent" }, "definition": { "modality": "text", "value": "A system that pursues a goal through multiple steps, often using tools or taking actions.", "short": "Multi-step tool user" }, "curatedDistractors": [ "Chatbot", "Model", "Dataset" ], "tags": [ "agent", "tool-use", "vocabulary" ], "uid": "tlursn1nyufxd" }, { "id": "ai-def-job-exposure", "shape": "definition", "term": { "modality": "text", "value": "Job exposure" }, "definition": { "modality": "text", "value": "The share of a job's tasks that AI could affect, assist, or automate—not a prediction that the job disappears.", "short": "Tasks AI could affect" }, "curatedDistractors": [ "Job displacement", "Task automation", "Labor shortage" ], "tags": [ "jobs", "automation", "vocabulary" ], "uid": "1pigfrz167pgd5" }, { "id": "ai-cloze-dartmouth", "shape": "cloze", "template": "The 1956 Dartmouth workshop helped name the field ___.", "answer": "artificial intelligence", "distractors": [ "machine learning", "cybernetics", "data science" ], "explanation": "The Dartmouth proposal used the term “artificial intelligence” for the new research field.", "tags": [ "dartmouth", "history", "vocabulary" ], "uid": "6uggmi1p0mwcy" }, { "id": "ai-cloze-parameters", "shape": "cloze", "template": "Learned internal values adjusted during training are called ___.", "answer": "parameters", "distractors": [ "datasets", "tokens", "prompts" ], "explanation": "Training adjusts parameters; data supplies the examples used to make those adjustments.", "tags": [ "parameters", "training", "vocabulary" ], "uid": "znkjok1h13jj0" }, { "id": "ai-cloze-fair-use", "shape": "cloze", "template": "U.S. courts apply the ___ doctrine through four statutory factors and case-specific facts.", "answer": "fair use", "distractors": [ "strict liability", "patent exhaustion", "trade secrecy" ], "explanation": "Fair use is a context-dependent copyright doctrine, not a blanket permission.", "tags": [ "copyright", "fair-use", "law" ], "uid": "ae93bt1okdp3n" }, { "id": "ai-proc-classify-phase", "shape": "procedure", "goal": "Classify an AI activity as training or inference", "steps": [ "Ask whether the model's learned weights are being changed", "If examples or feedback change weights, classify the activity as training", "If fixed weights process a new request, classify the activity as inference", "Separate retrieval or tool calls from changes to the model itself", "State the evidence for the classification and any uncertainty" ], "notes": "A deployed system may retrieve data or call tools during inference without retraining its underlying model.", "tags": [ "training", "inference", "procedure" ], "uid": "183iv23ucn3r5" }, { "id": "ai-proc-verify-output", "shape": "procedure", "goal": "Verify an important AI-generated output", "steps": [ "Identify the claims or decisions that matter", "Locate independent primary or authoritative sources", "Check each material claim against those sources", "Test calculations, citations, and edge cases separately", "Record uncertainty and decide whether human approval is required" ], "notes": "Polish and confidence are not substitutes for independent evidence.", "tags": [ "reliability", "verification", "procedure" ], "uid": "1hn39er1dbqmh9" }, { "id": "ot-cov-research-learning-1", "shape": "mcq", "prompt": { "modality": "text", "value": "Why name the specific thinking move (e.g., critique, reframe) you want AI to do?" }, "options": [ { "modality": "text", "value": "Removes the need to check citations afterward" }, { "modality": "text", "value": "Guarantees the AI never makes a factual error" }, { "modality": "text", "value": "Lets the AI decide your final values for you" }, { "modality": "text", "value": "Gives the model a clearer role you can judge" } ], "correctIndex": 3, "explanation": "The guide says naming the move gives the model a clearer role and helps you judge whether it was useful.", "tags": [ "thinking-moves", "ai-collaboration" ], "uid": "1ne4380e9so2o" }, { "id": "ot-cov-research-learning-2", "shape": "mcq", "prompt": { "modality": "text", "value": "Why keep brainstorming separate from judging ideas early on?" }, "options": [ { "modality": "text", "value": "Early judging can narrow the search too soon" }, { "modality": "text", "value": "Expansion and selection are one identical task" }, { "modality": "text", "value": "Selection criteria are invented after listing ends" }, { "modality": "text", "value": "Brainstorming alone produces a final decision" } ], "correctIndex": 0, "explanation": "The guide warns that judging while brainstorming can narrow the search too early; generate first, then apply criteria.", "tags": [ "thinking-moves", "decision-making" ], "uid": "17l79mp6xhhab" } ], "lessons": [ { "id": "afe-l1-prompt-anatomy", "title": "A Good Prompt Has Parts", "order": 0, "studyGuidePath": "/packs/ai-for-everyone/guides/l1-prompt-anatomy.md", "sources": [ { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" } ], "parts": [ { "id": "afe-l1p1-name-the-task", "title": "Name the Task", "order": 0, "blurb": "Turn a vague wish into a concrete action and outcome.", "studyGuideAnchor": "name-the-task", "itemIds": [ "afe-f-name-the-task-1", "afe-f-name-the-task-2", "afe-f-name-the-task-3", "afe-p-name-the-task-1", "afe-p-name-the-task-2", "afe-d-name-the-task", "afe-q-name-the-task-1", "afe-q-name-the-task-2", "afe-cmp-name-the-task", "tf-t-ai-for-everyone-afe-p-name-the-task-1" ] }, { "id": "afe-l1p2-give-context", "title": "Give Relevant Context", "order": 1, "blurb": "Supply the audience, situation, and material that change the answer.", "studyGuideAnchor": "give-relevant-context", "itemIds": [ "afe-f-give-context-1", "afe-f-give-context-2", "afe-f-give-context-3", "afe-p-give-context-1", "afe-p-give-context-2", "afe-d-give-context", "afe-q-give-context-1", "afe-q-give-context-2", "afe-cmp-give-context", "czr-ai-for-everyone-afe-d-give-context" ] }, { "id": "afe-l1p3-set-constraints", "title": "Set Constraints and Format", "order": 2, "blurb": "Specify boundaries that make the result usable.", "studyGuideAnchor": "set-constraints-and-format", "itemIds": [ "afe-f-set-constraints-1", "afe-f-set-constraints-2", "afe-f-set-constraints-3", "afe-p-set-constraints-1", "afe-p-set-constraints-2", "afe-d-set-constraints", "afe-q-set-constraints-1", "afe-q-set-constraints-2", "afe-cmp-set-constraints", "tf-t-ai-for-everyone-afe-p-set-constraints-1" ] }, { "id": "afe-l1p4-define-success", "title": "Define Success", "order": 3, "blurb": "Use examples and acceptance criteria to make quality visible.", "studyGuideAnchor": "define-success", "itemIds": [ "afe-f-define-success-1", "afe-f-define-success-2", "afe-f-define-success-3", "afe-p-define-success-1", "afe-p-define-success-2", "afe-d-define-success", "afe-q-define-success-1", "afe-q-define-success-2", "afe-cmp-define-success", "czr-ai-for-everyone-afe-d-define-success" ] } ], "objectives": [ { "id": "afe-obj-prompt-anatomy", "statement": "Build a prompt with a clear task, relevant context, useful constraints, and a testable output.", "demonstrationIds": [ "afe-demo-prompt-anatomy" ] } ], "demonstrations": [ { "id": "afe-demo-prompt-anatomy", "label": "A Good Prompt Has Parts practice", "itemIds": [ "afe-q-name-the-task-1", "afe-q-name-the-task-2", "afe-q-give-context-1", "afe-q-give-context-2", "afe-q-set-constraints-1", "afe-q-define-success-1", "afe-q-define-success-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-prompt-anatomy", "objectiveId": "afe-obj-prompt-anatomy", "title": "A Good Prompt Has Parts", "mcqIds": [ "afe-q-name-the-task-1", "afe-q-name-the-task-2", "afe-q-give-context-1", "afe-q-give-context-2", "afe-q-set-constraints-1" ] } ] }, { "id": "afe-l2-learn-by-showing", "title": "Learn by Showing", "order": 1, "studyGuidePath": "/packs/ai-for-everyone/guides/l2-learn-by-showing.md", "sources": [ { "label": "Min et al. — Rethinking the Role of Demonstrations", "url": "https://arxiv.org/abs/2202.12837" }, { "label": "Wang et al. — In-Context Learning from Invalid Demonstrations", "url": "https://aclanthology.org/2023.acl-long.153/" }, { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" } ], "parts": [ { "id": "afe-l2p1-zero-shot", "title": "Zero-Shot Requests", "order": 0, "blurb": "Ask directly when the task and output are already clear.", "studyGuideAnchor": "zero-shot-requests", "itemIds": [ "afe-f-zero-shot-1", "afe-f-zero-shot-2", "afe-f-zero-shot-3", "afe-p-zero-shot-1", "afe-p-zero-shot-2", "afe-d-zero-shot", "afe-q-zero-shot-1", "afe-q-zero-shot-2", "afe-cmp-zero-shot", "tf-t-ai-for-everyone-afe-p-zero-shot-1" ] }, { "id": "afe-l2p2-one-shot", "title": "One-Shot Examples", "order": 1, "blurb": "Show one compact model of the desired transformation.", "studyGuideAnchor": "one-shot-examples", "itemIds": [ "afe-f-one-shot-1", "afe-f-one-shot-2", "afe-f-one-shot-3", "afe-p-one-shot-1", "afe-p-one-shot-2", "afe-d-one-shot", "afe-q-one-shot-1", "afe-q-one-shot-2", "afe-cmp-one-shot", "czr-ai-for-everyone-afe-d-one-shot" ] }, { "id": "afe-l2p3-few-shot", "title": "Few-Shot Boundaries", "order": 2, "blurb": "Use several varied examples when a category has edges.", "studyGuideAnchor": "few-shot-boundaries", "itemIds": [ "afe-f-few-shot-1", "afe-f-few-shot-2", "afe-f-few-shot-3", "afe-p-few-shot-1", "afe-p-few-shot-2", "afe-d-few-shot", "afe-q-few-shot-1", "afe-q-few-shot-2", "afe-proc-few-shot", "tf-t-ai-for-everyone-afe-p-few-shot-1" ] }, { "id": "afe-l2p4-iterate", "title": "Iterate with Evidence", "order": 3, "blurb": "Change one thing, compare outputs, and keep what works.", "studyGuideAnchor": "iterate-with-evidence", "itemIds": [ "afe-f-iterate-1", "afe-f-iterate-2", "afe-f-iterate-3", "afe-p-iterate-1", "afe-p-iterate-2", "afe-d-iterate", "afe-q-iterate-1", "afe-q-iterate-2", "afe-proc-iterate", "czr-ai-for-everyone-afe-d-iterate" ] } ], "objectives": [ { "id": "afe-obj-learn-by-showing", "statement": "Choose zero-, one-, or few-shot prompting and improve examples through deliberate iteration.", "demonstrationIds": [ "afe-demo-learn-by-showing" ] } ], "demonstrations": [ { "id": "afe-demo-learn-by-showing", "label": "Learn by Showing practice", "itemIds": [ "afe-q-zero-shot-1", "afe-q-zero-shot-2", "afe-q-one-shot-1", "afe-q-one-shot-2", "afe-q-few-shot-1", "afe-q-iterate-1", "afe-q-iterate-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-learn-by-showing", "objectiveId": "afe-obj-learn-by-showing", "title": "Learn by Showing", "mcqIds": [ "afe-q-zero-shot-1", "afe-q-zero-shot-2", "afe-q-one-shot-1", "afe-q-one-shot-2", "afe-q-few-shot-1" ] } ] }, { "id": "afe-l3-roles-reasoning", "title": "Roles, Reasons, and Control", "order": 2, "studyGuidePath": "/packs/ai-for-everyone/guides/l3-roles-reasoning.md", "sources": [ { "label": "Zheng et al. — Role-Playing Does Not Improve Factual Accuracy", "url": "https://aclanthology.org/2024.findings-emnlp.888/" }, { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" } ], "parts": [ { "id": "afe-l3p1-role-prompts", "title": "Roles Are Lenses", "order": 0, "blurb": "Use a role to set perspective and standards—not to summon expertise.", "studyGuideAnchor": "roles-are-lenses", "itemIds": [ "afe-f-role-prompts-1", "afe-f-role-prompts-2", "afe-f-role-prompts-3", "afe-p-role-prompts-1", "afe-p-role-prompts-2", "afe-d-role-prompts", "afe-q-role-prompts-1", "afe-q-role-prompts-2", "afe-cmp-role-prompts", "tf-f-ai-for-everyone-afe-p-role-prompts-1" ] }, { "id": "afe-l3p2-visible-reasons", "title": "Ask for Visible Reasons", "order": 1, "blurb": "Request evidence, assumptions, and concise justification you can inspect.", "studyGuideAnchor": "ask-for-visible-reasons", "itemIds": [ "afe-f-visible-reasons-1", "afe-f-visible-reasons-2", "afe-f-visible-reasons-3", "afe-p-visible-reasons-1", "afe-p-visible-reasons-2", "afe-d-visible-reasons", "afe-q-visible-reasons-1", "afe-q-visible-reasons-2", "afe-cmp-visible-reasons", "czr-ai-for-everyone-afe-d-visible-reasons" ] }, { "id": "afe-l3p3-decompose", "title": "Break Work into Stages", "order": 2, "blurb": "Separate planning, production, and review when the task is complex.", "studyGuideAnchor": "break-work-into-stages", "itemIds": [ "afe-f-decompose-1", "afe-f-decompose-2", "afe-f-decompose-3", "afe-p-decompose-1", "afe-p-decompose-2", "afe-d-decompose", "afe-q-decompose-1", "afe-q-decompose-2", "afe-proc-decompose", "tf-t-ai-for-everyone-afe-p-decompose-1" ] }, { "id": "afe-l3p4-surface-uncertainty", "title": "Surface Assumptions and Uncertainty", "order": 3, "blurb": "Make missing information visible before it becomes invented detail.", "studyGuideAnchor": "surface-assumptions-and-uncertainty", "itemIds": [ "afe-f-surface-uncertainty-1", "afe-f-surface-uncertainty-2", "afe-f-surface-uncertainty-3", "afe-p-surface-uncertainty-1", "afe-p-surface-uncertainty-2", "afe-d-surface-uncertainty", "afe-q-surface-uncertainty-1", "afe-q-surface-uncertainty-2", "afe-proc-surface-uncertainty", "czr-ai-for-everyone-afe-d-surface-uncertainty" ] } ], "objectives": [ { "id": "afe-obj-roles-reasoning", "statement": "Use roles, decomposition, assumptions, and visible checks without mistaking performance cues for expertise.", "demonstrationIds": [ "afe-demo-roles-reasoning" ] } ], "demonstrations": [ { "id": "afe-demo-roles-reasoning", "label": "Roles, Reasons, and Control practice", "itemIds": [ "afe-q-role-prompts-1", "afe-q-role-prompts-2", "afe-q-visible-reasons-1", "afe-q-visible-reasons-2", "afe-q-decompose-1", "afe-q-surface-uncertainty-1", "afe-q-surface-uncertainty-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-roles-reasoning", "objectiveId": "afe-obj-roles-reasoning", "title": "Roles, Reasons, and Control", "mcqIds": [ "afe-q-role-prompts-1", "afe-q-role-prompts-2", "afe-q-visible-reasons-1", "afe-q-visible-reasons-2", "afe-q-decompose-1" ] } ] }, { "id": "afe-l4-agents", "title": "From Chatbots to Agents", "order": 3, "studyGuidePath": "/packs/ai-for-everyone/guides/l4-agents.md", "sources": [ { "label": "OpenAI — A practical guide to building agents", "url": "https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/" }, { "label": "NIST — Strengthening AI agent hijacking evaluations", "url": "https://www.nist.gov/news-events/news/2025/01/technical-blog-strengthening-ai-agent-hijacking-evaluations" }, { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" } ], "parts": [ { "id": "afe-l4p1-agent-loop", "title": "The Agent Loop", "order": 0, "blurb": "Understand how goals, actions, observations, and stopping fit together.", "studyGuideAnchor": "the-agent-loop", "itemIds": [ "afe-f-agent-loop-1", "afe-f-agent-loop-2", "afe-f-agent-loop-3", "afe-p-agent-loop-1", "afe-p-agent-loop-2", "afe-d-agent-loop", "afe-q-agent-loop-1", "afe-q-agent-loop-2", "afe-proc-agent-loop", "tf-f-ai-for-everyone-afe-p-agent-loop-1" ] }, { "id": "afe-l4p2-tools-permissions", "title": "Tools, State, and Permissions", "order": 1, "blurb": "Give an agent only the capabilities its task requires.", "studyGuideAnchor": "tools-state-and-permissions", "itemIds": [ "afe-f-tools-permissions-1", "afe-f-tools-permissions-2", "afe-f-tools-permissions-3", "afe-p-tools-permissions-1", "afe-p-tools-permissions-2", "afe-d-tools-permissions", "afe-q-tools-permissions-1", "afe-q-tools-permissions-2", "afe-proc-tools-permissions", "czr-ai-for-everyone-afe-d-tools-permissions" ] }, { "id": "afe-l4p3-approval-checkpoints", "title": "Human Approval Checkpoints", "order": 2, "blurb": "Place review where errors become expensive or hard to reverse.", "studyGuideAnchor": "human-approval-checkpoints", "itemIds": [ "afe-f-approval-checkpoints-1", "afe-f-approval-checkpoints-2", "afe-f-approval-checkpoints-3", "afe-p-approval-checkpoints-1", "afe-p-approval-checkpoints-2", "afe-d-approval-checkpoints", "afe-q-approval-checkpoints-1", "afe-q-approval-checkpoints-2", "afe-cmp-approval-checkpoints", "tf-t-ai-for-everyone-afe-p-approval-checkpoints-1" ] }, { "id": "afe-l4p4-agent-failures", "title": "Agent Failure Modes", "order": 3, "blurb": "Expect drift, tool errors, hostile instructions, and silent loops.", "studyGuideAnchor": "agent-failure-modes", "itemIds": [ "afe-f-agent-failures-1", "afe-f-agent-failures-2", "afe-f-agent-failures-3", "afe-p-agent-failures-1", "afe-p-agent-failures-2", "afe-d-agent-failures", "afe-q-agent-failures-1", "afe-q-agent-failures-2", "afe-proc-agent-failures", "czr-ai-for-everyone-afe-d-agent-failures" ] } ], "objectives": [ { "id": "afe-obj-agents", "statement": "Design bounded agent workflows with tools, permissions, checkpoints, and explicit stopping conditions.", "demonstrationIds": [ "afe-demo-agents" ] } ], "demonstrations": [ { "id": "afe-demo-agents", "label": "From Chatbots to Agents practice", "itemIds": [ "afe-q-agent-loop-1", "afe-q-agent-loop-2", "afe-q-tools-permissions-1", "afe-q-tools-permissions-2", "afe-q-approval-checkpoints-1" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-agents", "objectiveId": "afe-obj-agents", "title": "From Chatbots to Agents", "mcqIds": [ "afe-q-agent-loop-1", "afe-q-agent-loop-2", "afe-q-tools-permissions-1", "afe-q-tools-permissions-2", "afe-q-approval-checkpoints-1" ] } ] }, { "id": "afe-l5-multimodal", "title": "Make with Multiple Media", "order": 4, "studyGuidePath": "/packs/ai-for-everyone/guides/l5-multimodal.md", "sources": [ { "label": "OpenAI Academy — Image generation", "url": "https://openai.com/academy/image-generation/" }, { "label": "Google Cloud — Configure image aspect ratio", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/image/configure-aspect-ratio" }, { "label": "Rombach et al. — High-Resolution Image Synthesis with Latent Diffusion Models", "url": "https://arxiv.org/abs/2112.10752" }, { "label": "Radford et al. — Robust Speech Recognition via Large-Scale Weak Supervision", "url": "https://arxiv.org/abs/2212.04356" }, { "label": "FTC — Approaches to address AI-enabled voice cloning", "url": "https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2024/04/approaches-address-ai-enabled-voice-cloning" } ], "parts": [ { "id": "afe-l5p1-image-prompts", "title": "Prompt an Image", "order": 0, "blurb": "Describe subject, setting, composition, light, style, and exclusions.", "studyGuideAnchor": "prompt-an-image", "itemIds": [ "afe-f-image-prompts-1", "afe-f-image-prompts-2", "afe-f-image-prompts-3", "afe-p-image-prompts-1", "afe-p-image-prompts-2", "afe-d-image-prompts", "afe-q-image-prompts-1", "afe-q-image-prompts-2", "afe-proc-image-prompts", "tf-t-ai-for-everyone-afe-p-image-prompts-1" ] }, { "id": "afe-l5p2-aspect-ratios", "title": "Choose the Canvas", "order": 1, "blurb": "Match aspect ratio and composition to where the image will live.", "studyGuideAnchor": "choose-the-canvas", "itemIds": [ "afe-f-aspect-ratios-1", "afe-f-aspect-ratios-2", "afe-f-aspect-ratios-3", "afe-n-aspect-ratios-1", "afe-n-aspect-ratios-2", "afe-n-aspect-ratios-3", "afe-n-aspect-ratios-4", "afe-n-aspect-ratios-5", "afe-n-aspect-ratios-6", "afe-q-aspect-ratios-special" ] }, { "id": "afe-l5p3-video-planning", "title": "Plan Video as Shots", "order": 2, "blurb": "Build a sequence with beats, continuity, and edit points.", "studyGuideAnchor": "plan-video-as-shots", "itemIds": [ "afe-f-video-planning-1", "afe-f-video-planning-2", "afe-f-video-planning-3", "afe-p-video-planning-1", "afe-p-video-planning-2", "afe-d-video-planning", "afe-q-video-planning-1", "afe-q-video-planning-2", "afe-proc-video-planning", "czr-ai-for-everyone-afe-d-video-planning" ] }, { "id": "afe-l5p4-audio-workflows", "title": "Work with Audio", "order": 3, "blurb": "Use speech tools while preserving consent, identity, and review.", "studyGuideAnchor": "work-with-audio", "itemIds": [ "afe-f-audio-workflows-1", "afe-f-audio-workflows-2", "afe-f-audio-workflows-3", "afe-k-audio-workflows-1", "afe-k-audio-workflows-2", "afe-k-audio-workflows-3", "afe-k-audio-workflows-4", "afe-k-audio-workflows-5", "afe-k-audio-workflows-6", "afe-q-audio-workflows-special" ] } ], "objectives": [ { "id": "afe-obj-multimodal", "statement": "Prompt and review image, video, and audio systems with attention to composition, continuity, consent, and fit.", "demonstrationIds": [ "afe-demo-multimodal" ] } ], "demonstrations": [ { "id": "afe-demo-multimodal", "label": "Make with Multiple Media practice", "itemIds": [ "afe-q-image-prompts-1", "afe-q-image-prompts-2", "afe-q-aspect-ratios-special", "afe-q-video-planning-1", "afe-q-video-planning-2", "afe-q-audio-workflows-special" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-multimodal", "objectiveId": "afe-obj-multimodal", "title": "Make with Multiple Media", "mcqIds": [ "afe-q-image-prompts-1", "afe-q-image-prompts-2", "afe-q-aspect-ratios-special", "afe-q-video-planning-1", "afe-q-video-planning-2" ] } ] }, { "id": "afe-l6-writing-documents", "title": "Writing and Documents", "order": 5, "studyGuidePath": "/packs/ai-for-everyone/guides/l6-writing-documents.md", "sources": [ { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" }, { "label": "Liu et al. — Lost in the Middle", "url": "https://aclanthology.org/2024.tacl-1.9/" }, { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" } ], "parts": [ { "id": "afe-l6p1-draft-edit", "title": "Separate Drafting from Editing", "order": 0, "blurb": "Choose whether the system should create, diagnose, or revise.", "studyGuideAnchor": "separate-drafting-from-editing", "itemIds": [ "afe-f-draft-edit-1", "afe-f-draft-edit-2", "afe-f-draft-edit-3", "afe-p-draft-edit-1", "afe-p-draft-edit-2", "afe-d-draft-edit", "afe-q-draft-edit-1", "afe-q-draft-edit-2", "afe-p-draft-edit-3", "tf-f-ai-for-everyone-afe-p-draft-edit-1" ] }, { "id": "afe-l6p2-summarize", "title": "Summarize Long Material", "order": 1, "blurb": "Define the audience, coverage, and evidence for a summary.", "studyGuideAnchor": "summarize-long-material", "itemIds": [ "afe-f-summarize-1", "afe-f-summarize-2", "afe-f-summarize-3", "afe-p-summarize-1", "afe-p-summarize-2", "afe-d-summarize", "afe-q-summarize-1", "afe-q-summarize-2", "afe-cmp-summarize", "czr-ai-for-everyone-afe-d-summarize" ] }, { "id": "afe-l6p3-transform", "title": "Transform Tone and Format", "order": 2, "blurb": "Change the container without accidentally changing the claim.", "studyGuideAnchor": "transform-tone-and-format", "itemIds": [ "afe-f-transform-1", "afe-f-transform-2", "afe-f-transform-3", "afe-p-transform-1", "afe-p-transform-2", "afe-d-transform", "afe-q-transform-1", "afe-q-transform-2", "afe-p-transform-3", "tf-f-ai-for-everyone-afe-p-transform-1" ] }, { "id": "afe-l6p4-document-qa", "title": "Ask Questions of Documents", "order": 3, "blurb": "Extract structured answers and distinguish absence from evidence.", "studyGuideAnchor": "ask-questions-of-documents", "itemIds": [ "afe-f-document-qa-1", "afe-f-document-qa-2", "afe-f-document-qa-3", "afe-p-document-qa-1", "afe-p-document-qa-2", "afe-d-document-qa", "afe-q-document-qa-1", "afe-q-document-qa-2", "afe-p-document-qa-3", "czr-ai-for-everyone-afe-d-document-qa" ] } ], "objectives": [ { "id": "afe-obj-writing-documents", "statement": "Use AI to draft, revise, summarize, transform, and inspect documents while preserving meaning and provenance.", "demonstrationIds": [ "afe-demo-writing-documents" ] } ], "demonstrations": [ { "id": "afe-demo-writing-documents", "label": "Writing and Documents practice", "itemIds": [ "afe-q-draft-edit-1", "afe-q-draft-edit-2", "afe-q-summarize-1", "afe-q-summarize-2", "afe-q-transform-1", "afe-q-document-qa-1", "afe-q-document-qa-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-writing-documents", "objectiveId": "afe-obj-writing-documents", "title": "Writing and Documents", "mcqIds": [ "afe-q-draft-edit-1", "afe-q-draft-edit-2", "afe-q-summarize-1", "afe-q-summarize-2", "afe-q-transform-1" ] } ] }, { "id": "afe-l7-research-learning", "title": "Research, Learning, and Decisions", "order": 6, "studyGuidePath": "/packs/ai-for-everyone/guides/l7-research-learning.md", "sources": [ { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, { "label": "UNESCO — Guidance for generative AI in education and research", "url": "https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research" }, { "label": "Liu et al. — Lost in the Middle", "url": "https://aclanthology.org/2024.tacl-1.9/" } ], "parts": [ { "id": "afe-l7p1-research-plan", "title": "Plan Research Before Answering", "order": 0, "blurb": "Turn a broad question into claims, evidence needs, and a search path.", "studyGuideAnchor": "plan-research-before-answering", "itemIds": [ "afe-f-research-plan-1", "afe-f-research-plan-2", "afe-f-research-plan-3", "afe-p-research-plan-1", "afe-p-research-plan-2", "afe-d-research-plan", "afe-q-research-plan-1", "afe-q-research-plan-2", "afe-p-research-plan-3", "tf-t-ai-for-everyone-afe-p-research-plan-1" ] }, { "id": "afe-l7p2-tutor", "title": "Use AI as a Practice Partner", "order": 1, "blurb": "Ask for questions, hints, feedback, and retrieval—not only answers.", "studyGuideAnchor": "use-ai-as-a-practice-partner", "itemIds": [ "afe-f-tutor-1", "afe-f-tutor-2", "afe-f-tutor-3", "afe-p-tutor-1", "afe-p-tutor-2", "afe-d-tutor", "afe-q-tutor-1", "afe-q-tutor-2", "afe-p-tutor-3", "czr-ai-for-everyone-afe-d-tutor" ] }, { "id": "afe-l7p3-decision-support", "title": "Compare Options without Outsourcing Choice", "order": 2, "blurb": "Make criteria, tradeoffs, and uncertainty explicit.", "studyGuideAnchor": "compare-options-without-outsourcing-choice", "itemIds": [ "afe-f-decision-support-1", "afe-f-decision-support-2", "afe-f-decision-support-3", "afe-p-decision-support-1", "afe-p-decision-support-2", "afe-d-decision-support", "afe-q-decision-support-1", "afe-q-decision-support-2", "afe-p-decision-support-3", "tf-f-ai-for-everyone-afe-p-decision-support-1" ] }, { "id": "afe-l7p4-thinking-moves", "title": "Six Thinking Moves", "order": 3, "blurb": "Choose the cognitive move that matches the moment.", "studyGuideAnchor": "six-thinking-moves", "itemIds": [ "afe-f-thinking-moves-1", "afe-f-thinking-moves-2", "afe-f-thinking-moves-3", "afe-k-thinking-moves-1", "afe-k-thinking-moves-2", "afe-k-thinking-moves-3", "afe-k-thinking-moves-4", "afe-k-thinking-moves-5", "afe-k-thinking-moves-6", "afe-q-thinking-moves-special", "ot-cov-research-learning-1", "ot-cov-research-learning-2" ] } ], "objectives": [ { "id": "afe-obj-research-learning", "statement": "Use AI to structure inquiry, practice learning, compare options, and perform distinct thinking moves without outsourcing judgment.", "demonstrationIds": [ "afe-demo-research-learning" ] } ], "demonstrations": [ { "id": "afe-demo-research-learning", "label": "Research, Learning, and Decisions practice", "itemIds": [ "afe-q-research-plan-1", "afe-q-research-plan-2", "afe-q-tutor-1", "afe-q-tutor-2", "afe-q-decision-support-1", "ot-cov-research-learning-1", "ot-cov-research-learning-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-research-learning", "objectiveId": "afe-obj-research-learning", "title": "Research, Learning, and Decisions", "mcqIds": [ "afe-q-research-plan-1", "afe-q-research-plan-2", "afe-q-tutor-1", "afe-q-tutor-2", "afe-q-decision-support-1", "ot-cov-research-learning-1", "ot-cov-research-learning-2" ] } ] }, { "id": "afe-l8-data-code", "title": "Data, Spreadsheets, and Code", "order": 7, "studyGuidePath": "/packs/ai-for-everyone/guides/l8-data-code.md", "sources": [ { "label": "NIST — Secure Software Development Framework", "url": "https://csrc.nist.gov/projects/ssdf" }, { "label": "Lian et al. — A.S.E. security benchmark for AI-generated code", "url": "https://aclanthology.org/2026.findings-acl.1569/" }, { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" } ], "parts": [ { "id": "afe-l8p1-prepare-data", "title": "Describe the Data First", "order": 0, "blurb": "Supply schema, units, meanings, and missing-value rules before analysis.", "studyGuideAnchor": "describe-the-data-first", "itemIds": [ "afe-f-prepare-data-1", "afe-f-prepare-data-2", "afe-f-prepare-data-3", "afe-p-prepare-data-1", "afe-p-prepare-data-2", "afe-d-prepare-data", "afe-q-prepare-data-1", "afe-q-prepare-data-2", "afe-p-prepare-data-3", "czr-ai-for-everyone-afe-d-prepare-data" ] }, { "id": "afe-l8p2-formulas-charts", "title": "Build Formulas and Charts", "order": 1, "blurb": "Ask for transparent calculations and verify them on small examples.", "studyGuideAnchor": "build-formulas-and-charts", "itemIds": [ "afe-f-formulas-charts-1", "afe-f-formulas-charts-2", "afe-f-formulas-charts-3", "afe-p-formulas-charts-1", "afe-p-formulas-charts-2", "afe-d-formulas-charts", "afe-q-formulas-charts-1", "afe-q-formulas-charts-2", "afe-p-formulas-charts-3", "tf-t-ai-for-everyone-afe-p-formulas-charts-1" ] }, { "id": "afe-l8p3-coding-cycle", "title": "Use a Coding Cycle", "order": 2, "blurb": "Inspect the repository, plan the change, patch narrowly, and test.", "studyGuideAnchor": "use-a-coding-cycle", "itemIds": [ "afe-f-coding-cycle-1", "afe-f-coding-cycle-2", "afe-f-coding-cycle-3", "afe-p-coding-cycle-1", "afe-p-coding-cycle-2", "afe-d-coding-cycle", "afe-q-coding-cycle-1", "afe-q-coding-cycle-2", "afe-p-coding-cycle-3", "czr-ai-for-everyone-afe-d-coding-cycle" ] }, { "id": "afe-l8p4-test-secure", "title": "Test and Secure Generated Code", "order": 3, "blurb": "Treat generated code as untrusted until reviewed and exercised.", "studyGuideAnchor": "test-and-secure-generated-code", "itemIds": [ "afe-f-test-secure-1", "afe-f-test-secure-2", "afe-f-test-secure-3", "afe-p-test-secure-1", "afe-p-test-secure-2", "afe-d-test-secure", "afe-q-test-secure-1", "afe-q-test-secure-2", "afe-p-test-secure-3", "tf-f-ai-for-everyone-afe-p-test-secure-1" ] } ], "objectives": [ { "id": "afe-obj-data-code", "statement": "Use AI to inspect structured data and code through schemas, tests, security checks, and reversible changes.", "demonstrationIds": [ "afe-demo-data-code" ] } ], "demonstrations": [ { "id": "afe-demo-data-code", "label": "Data, Spreadsheets, and Code practice", "itemIds": [ "afe-q-prepare-data-1", "afe-q-prepare-data-2", "afe-q-formulas-charts-1", "afe-q-formulas-charts-2", "afe-q-coding-cycle-1", "afe-q-test-secure-1", "afe-q-test-secure-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-data-code", "objectiveId": "afe-obj-data-code", "title": "Data, Spreadsheets, and Code", "mcqIds": [ "afe-q-prepare-data-1", "afe-q-prepare-data-2", "afe-q-formulas-charts-1", "afe-q-formulas-charts-2", "afe-q-coding-cycle-1" ] } ] }, { "id": "afe-l9-verify-protect", "title": "Verify and Protect", "order": 8, "studyGuidePath": "/packs/ai-for-everyone/guides/l9-verify-protect.md", "sources": [ { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, { "label": "Office of the Privacy Commissioner of Canada — AI chatbots", "url": "https://www.priv.gc.ca/en/privacy-topics/technology/artificial-intelligence/ai-chatbots_ind/" }, { "label": "FTC — Scammers use AI to enhance family-emergency schemes", "url": "https://consumer.ftc.gov/consumer-alerts/2023/03/scammers-use-ai-enhance-their-family-emergency-schemes" }, { "label": "NIST — Strengthening AI agent hijacking evaluations", "url": "https://www.nist.gov/news-events/news/2025/01/technical-blog-strengthening-ai-agent-hijacking-evaluations" } ], "parts": [ { "id": "afe-l9p1-verify-claims", "title": "Verify Claims and Citations", "order": 0, "blurb": "Break important output into checkable claims and trace each one.", "studyGuideAnchor": "verify-claims-and-citations", "itemIds": [ "afe-f-verify-claims-1", "afe-f-verify-claims-2", "afe-f-verify-claims-3", "afe-p-verify-claims-1", "afe-p-verify-claims-2", "afe-d-verify-claims", "afe-q-verify-claims-1", "afe-q-verify-claims-2", "afe-cmp-verify-claims", "czr-ai-for-everyone-afe-d-verify-claims" ] }, { "id": "afe-l9p2-privacy", "title": "Minimize Sensitive Data", "order": 1, "blurb": "Share only what the task needs and understand where it goes.", "studyGuideAnchor": "minimize-sensitive-data", "itemIds": [ "afe-f-privacy-1", "afe-f-privacy-2", "afe-f-privacy-3", "afe-p-privacy-1", "afe-p-privacy-2", "afe-d-privacy", "afe-q-privacy-1", "afe-q-privacy-2", "afe-proc-privacy", "tf-f-ai-for-everyone-afe-p-privacy-1" ] }, { "id": "afe-l9p3-manipulation", "title": "Resist Injection and Impersonation", "order": 2, "blurb": "Treat external instructions, urgent voices, and polished messages as untrusted.", "studyGuideAnchor": "resist-injection-and-impersonation", "itemIds": [ "afe-f-manipulation-1", "afe-f-manipulation-2", "afe-f-manipulation-3", "afe-p-manipulation-1", "afe-p-manipulation-2", "afe-d-manipulation", "afe-q-manipulation-1", "afe-q-manipulation-2", "afe-p-manipulation-3", "czr-ai-for-everyone-afe-d-manipulation" ] }, { "id": "afe-l9p4-high-stakes", "title": "Know the High-Stakes Boundary", "order": 3, "blurb": "Use AI to prepare questions and organize evidence—not to replace qualified judgment.", "studyGuideAnchor": "know-the-high-stakes-boundary", "itemIds": [ "afe-f-high-stakes-1", "afe-f-high-stakes-2", "afe-f-high-stakes-3", "afe-p-high-stakes-1", "afe-p-high-stakes-2", "afe-d-high-stakes", "afe-q-high-stakes-1", "afe-q-high-stakes-2", "afe-p-high-stakes-3", "tf-f-ai-for-everyone-afe-p-high-stakes-1" ] } ], "objectives": [ { "id": "afe-obj-verify-protect", "statement": "Verify material claims, minimize sensitive data, resist manipulation, and keep high-stakes decisions with qualified people.", "demonstrationIds": [ "afe-demo-verify-protect" ] } ], "demonstrations": [ { "id": "afe-demo-verify-protect", "label": "Verify and Protect practice", "itemIds": [ "afe-q-verify-claims-1", "afe-q-verify-claims-2", "afe-q-privacy-1", "afe-q-privacy-2", "afe-q-manipulation-1", "afe-q-high-stakes-1", "afe-q-high-stakes-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-verify-protect", "objectiveId": "afe-obj-verify-protect", "title": "Verify and Protect", "mcqIds": [ "afe-q-verify-claims-1", "afe-q-verify-claims-2", "afe-q-privacy-1", "afe-q-privacy-2", "afe-q-manipulation-1" ] } ] }, { "id": "afe-l10-personal-workflow", "title": "Build Your Personal AI Workflow", "order": 9, "studyGuidePath": "/packs/ai-for-everyone/guides/l10-personal-workflow.md", "sources": [ { "label": "NIST — AI Risk Management Framework Playbook", "url": "https://airc.nist.gov/airmf-resources/playbook/" }, { "label": "NIST — Generative AI Profile", "url": "https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence" }, { "label": "Google Cloud — Prompt design strategies", "url": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies" } ], "parts": [ { "id": "afe-l10p1-choose-tools", "title": "Choose the Tool for the Task", "order": 0, "blurb": "Compare capability, data handling, control, cost, and reversibility.", "studyGuideAnchor": "choose-the-tool-for-the-task", "itemIds": [ "afe-f-choose-tools-1", "afe-f-choose-tools-2", "afe-f-choose-tools-3", "afe-p-choose-tools-1", "afe-p-choose-tools-2", "afe-d-choose-tools", "afe-q-choose-tools-1", "afe-q-choose-tools-2", "afe-cmp-choose-tools", "czr-ai-for-everyone-afe-d-choose-tools" ] }, { "id": "afe-l10p2-templates", "title": "Turn Success into a Template", "order": 1, "blurb": "Save the structure, variables, example, and review checklist.", "studyGuideAnchor": "turn-success-into-a-template", "itemIds": [ "afe-f-templates-1", "afe-f-templates-2", "afe-f-templates-3", "afe-p-templates-1", "afe-p-templates-2", "afe-d-templates", "afe-q-templates-1", "afe-q-templates-2", "afe-p-templates-3", "tf-t-ai-for-everyone-afe-p-templates-1" ] }, { "id": "afe-l10p3-evaluate", "title": "Use a Small Evaluation Set", "order": 2, "blurb": "Score quality, failure severity, time, and cost on real cases.", "studyGuideAnchor": "use-a-small-evaluation-set", "itemIds": [ "afe-f-evaluate-1", "afe-f-evaluate-2", "afe-f-evaluate-3", "afe-p-evaluate-1", "afe-p-evaluate-2", "afe-d-evaluate", "afe-q-evaluate-1", "afe-q-evaluate-2", "afe-p-evaluate-3", "czr-ai-for-everyone-afe-d-evaluate" ] }, { "id": "afe-l10p4-handoffs", "title": "Practice, Then Go Deeper", "order": 3, "blurb": "Use a bounded personal pilot and the rest of the Learn AI collection.", "studyGuideAnchor": "practice-then-go-deeper", "itemIds": [ "afe-f-handoffs-1", "afe-f-handoffs-2", "afe-p-handoffs-1", "afe-p-handoffs-2", "afe-d-handoffs", "afe-q-handoffs-1", "afe-q-handoffs-2", "afe-p-handoffs-3", "tf-t-ai-for-everyone-afe-p-handoffs-1" ] } ], "objectives": [ { "id": "afe-obj-personal-workflow", "statement": "Choose tools deliberately, turn successful prompts into templates, evaluate them on real cases, and route deeper learning through the collection.", "demonstrationIds": [ "afe-demo-personal-workflow" ] } ], "demonstrations": [ { "id": "afe-demo-personal-workflow", "label": "Build Your Personal AI Workflow practice", "itemIds": [ "afe-q-choose-tools-1", "afe-q-choose-tools-2", "afe-q-templates-1", "afe-q-templates-2", "afe-q-evaluate-1", "afe-q-handoffs-1", "afe-q-handoffs-2" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "afe-test-personal-workflow", "objectiveId": "afe-obj-personal-workflow", "title": "Build Your Personal AI Workflow", "mcqIds": [ "afe-q-choose-tools-1", "afe-q-choose-tools-2", "afe-q-templates-1", "afe-q-templates-2", "afe-q-evaluate-1" ] } ] }, { "id": "ai-l1-map", "title": "A Map of AI", "order": 10, "studyGuidePath": "/packs/ai-for-everyone/guides/ai-l1-map.md", "sources": [ { "label": "Dartmouth — Artificial Intelligence coined at Dartmouth", "url": "https://home.dartmouth.edu/about/artificial-intelligence-ai-coined-dartmouth" }, { "label": "NIST — AI Risk Management Framework", "url": "https://www.nist.gov/itl/ai-risk-management-framework" }, { "label": "FlashFeed — LLM 101" } ], "parts": [ { "id": "ai-l1p1-intelligence", "title": "Narrow Systems, AGI, and Superintelligence", "order": 0, "blurb": "Separate present systems from hypothetical categories.", "studyGuideAnchor": "narrow-systems-agi-and-superintelligence", "itemIds": [ "ai-fact-narrow-systems", "ai-fact-agi-asi", "ai-cmp-fluency-general", "ai-pair-broad-fluency", "ai-mcq-alphago-boundary", "ai-mcq-superintelligence-status", "ai-def-agi" ] }, { "id": "ai-l1p2-phases", "title": "Training and Inference", "order": 1, "blurb": "Learn the two-phase distinction behind modern AI systems.", "studyGuideAnchor": "training-and-inference", "itemIds": [ "ai-fact-training", "ai-fact-inference", "ai-cmp-train-infer", "ai-pair-inference-weights", "ai-mcq-weights-change", "ai-mcq-query-phase", "ai-def-inference", "ai-proc-classify-phase" ] }, { "id": "ai-l1p3-history", "title": "Six Milestones in AI", "order": 2, "blurb": "Put the current wave on a timeline, then keep human projection in check.", "studyGuideAnchor": "six-milestones-in-ai", "itemIds": [ "ai-num-dartmouth", "ai-num-deep-blue", "ai-num-alexnet", "ai-num-alphago", "ai-num-transformer", "ai-num-chatgpt", "ai-mcq-ai-winter", "ai-pair-ai-winter", "ai-cloze-dartmouth" ] } ], "objectives": [ { "id": "ai-obj-map", "statement": "Distinguish present AI systems, hypothetical intelligence categories, and the training and inference phases.", "demonstrationIds": [ "ai-demo-map" ] } ], "demonstrations": [ { "id": "ai-demo-map", "label": "AI categories and phases", "itemIds": [ "ai-mcq-alphago-boundary", "ai-mcq-superintelligence-status", "ai-mcq-weights-change", "ai-mcq-query-phase", "ai-mcq-ai-winter" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "ai-test-map", "objectiveId": "ai-obj-map", "title": "The AI Map", "mcqIds": [ "ai-mcq-alphago-boundary", "ai-mcq-superintelligence-status", "ai-mcq-weights-change", "ai-mcq-query-phase", "ai-mcq-ai-winter" ] } ] }, { "id": "ai-l2-vocabulary", "title": "Words You'll See Everywhere", "order": 11, "studyGuidePath": "/packs/ai-for-everyone/guides/ai-l2-vocabulary.md", "sources": [ { "label": "Google Cloud — What are AI hallucinations?", "url": "https://cloud.google.com/discover/what-are-ai-hallucinations" }, { "label": "NIST — AI Risk Management Framework", "url": "https://www.nist.gov/itl/ai-risk-management-framework" }, { "label": "Wang et al. — Survey of LLM-based agents", "url": "https://arxiv.org/abs/2309.07864" } ], "parts": [ { "id": "ai-l2p1-building-blocks", "title": "Models, Data, and Parameters", "order": 0, "blurb": "Name the basic pieces without diving into architecture.", "studyGuideAnchor": "models-data-and-parameters", "itemIds": [ "ai-fact-model", "ai-fact-data-parameters", "ai-cmp-data-parameters", "ai-pair-parameter-count", "ai-mcq-dataset-role", "ai-mcq-parameter-role", "ai-def-parameters", "ai-cloze-parameters" ] }, { "id": "ai-l2p2-failure-modes", "title": "Hallucination, Bias, and Grounding", "order": 1, "blurb": "Tell fabrication from systematic distortion and learn the main mitigation.", "studyGuideAnchor": "hallucination-bias-and-grounding", "itemIds": [ "ai-fact-hallucination", "ai-fact-bias", "ai-cmp-hallucination-bias", "ai-pair-grounding-limits", "ai-mcq-invented-source", "ai-mcq-grounding", "ai-def-hallucination" ] }, { "id": "ai-l2p3-agents", "title": "Six Key AI Concepts", "order": 2, "blurb": "Recognize the system, interface, agent loop, and three common reasoning terms.", "studyGuideAnchor": "six-key-ai-concepts", "itemIds": [ "ai-concept-model", "ai-concept-chatbot", "ai-concept-agent", "ai-concept-hallucination", "ai-concept-grounding", "ai-concept-anthropomorphism", "ai-mcq-agent-addition", "ai-pair-agent-workflow", "ai-def-agent" ] } ], "objectives": [ { "id": "ai-obj-vocabulary", "statement": "Use collection-wide AI vocabulary to distinguish components, failure modes, interfaces, and agents.", "demonstrationIds": [ "ai-demo-vocabulary" ] } ], "demonstrations": [ { "id": "ai-demo-vocabulary", "label": "Core AI vocabulary", "itemIds": [ "ai-mcq-dataset-role", "ai-mcq-parameter-role", "ai-mcq-invented-source", "ai-mcq-grounding", "ai-mcq-agent-addition", "ai-concept-chatbot" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "ai-test-vocabulary", "objectiveId": "ai-obj-vocabulary", "title": "Shared Vocabulary", "mcqIds": [ "ai-mcq-dataset-role", "ai-mcq-parameter-role", "ai-mcq-invented-source", "ai-mcq-grounding", "ai-mcq-agent-addition" ] } ] }, { "id": "ai-l3-limits", "title": "What AI Can—and Cannot—Do", "order": 12, "studyGuidePath": "/packs/ai-for-everyone/guides/ai-l3-limits.md", "sources": [ { "label": "NIST — AI Risk Management Framework", "url": "https://www.nist.gov/itl/ai-risk-management-framework" }, { "label": "IEA — Energy and AI", "url": "https://www.iea.org/reports/energy-and-ai" }, { "label": "European Commission — AI Act framework", "url": "https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai" }, { "label": "White House — America's AI Action Plan", "url": "https://www.whitehouse.gov/srv/htdocs/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf" } ], "parts": [ { "id": "ai-l3p1-reliability", "title": "Strengths and the Reliability Gap", "order": 0, "blurb": "Match the task to the tool and verify what matters.", "studyGuideAnchor": "strengths-and-the-reliability-gap", "itemIds": [ "ai-fact-strengths", "ai-fact-reliability-gap", "ai-cmp-task-fit", "ai-pair-confident-tone", "ai-mcq-first-use", "ai-mcq-reliability-signal", "ai-proc-verify-output" ] }, { "id": "ai-l3p2-footprint", "title": "Energy, Water, and Physical Scale", "order": 1, "blurb": "Keep the footprint in context and the measurement boundary visible.", "studyGuideAnchor": "energy-water-and-physical-scale", "itemIds": [ "ai-fact-energy", "ai-fact-water-context", "ai-cmp-energy-claim", "ai-pair-physical-cost", "ai-mcq-footprint-context", "ai-mcq-iea-scope" ] }, { "id": "ai-l3p3-regulation", "title": "A Regulation Snapshot", "order": 2, "blurb": "Compare the EU and U.S. without pretending one rule is global.", "studyGuideAnchor": "a-regulation-snapshot", "itemIds": [ "ai-fact-eu-act", "ai-fact-us-policy", "ai-cmp-eu-us", "ai-pair-jurisdiction", "ai-mcq-eu-organizing-idea", "ai-def-ai-governance" ] } ], "objectives": [ { "id": "ai-obj-limits", "statement": "Choose reviewable AI uses, interpret footprint claims, and recognize jurisdiction-specific governance.", "demonstrationIds": [ "ai-demo-limits" ] } ], "demonstrations": [ { "id": "ai-demo-limits", "label": "Capabilities, footprint, and governance", "itemIds": [ "ai-mcq-first-use", "ai-mcq-reliability-signal", "ai-mcq-footprint-context", "ai-mcq-iea-scope", "ai-mcq-eu-organizing-idea" ], "requiredCorrect": 4 } ], "objectiveTests": [ { "id": "ai-test-limits", "objectiveId": "ai-obj-limits", "title": "Honest Limits", "mcqIds": [ "ai-mcq-first-use", "ai-mcq-reliability-signal", "ai-mcq-footprint-context", "ai-mcq-iea-scope", "ai-mcq-eu-organizing-idea" ] } ] }, { "id": "ai-l4-disagreements", "title": "Where Smart People Disagree", "order": 13, "studyGuidePath": "/packs/ai-for-everyone/guides/ai-l4-disagreements.md", "sources": [ { "label": "IMF — AI and the Future of Work", "url": "https://www.elibrary.imf.org/view/journals/006/2024/001/article-A001-en.xml" }, { "label": "Acemoglu — The Simple Macroeconomics of AI", "url": "https://economics.mit.edu/sites/default/files/2024-04/The%20Simple%20Macroeconomics%20of%20AI.pdf" }, { "label": "Forecasting Research Institute — XPT report", "url": "https://forecastingresearch.org/s/XPT.pdf" }, { "label": "U.S. Copyright Office — Generative AI Training", "url": "https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf" }, { "label": "The Copyright Society — Training on Trial", "url": "https://copyrightsociety.org/wp-content/uploads/2026/05/73-J.-Copyright-Socy-261-2026.pdf" }, { "label": "Grace et al. — Thousands of AI Authors", "url": "https://arxiv.org/abs/2401.02843" } ], "parts": [ { "id": "ai-l4p1-jobs", "title": "Will AI Take Your Job?", "order": 0, "blurb": "Separate task exposure, automation, adoption, and employment outcomes.", "studyGuideAnchor": "will-ai-take-your-job", "itemIds": [ "ai-fact-jobs-tasks", "ai-fact-job-estimates", "ai-cmp-exposure-displacement", "ai-pair-job-exposure", "ai-mcq-job-unit", "ai-mcq-estimate-conflict", "ai-def-job-exposure" ] }, { "id": "ai-l4p2-xrisk", "title": "Catastrophic and Existential Risk", "order": 1, "blurb": "Understand both the precautionary and skeptical cases.", "studyGuideAnchor": "catastrophic-and-existential-risk", "itemIds": [ "ai-fact-xrisk-concern", "ai-fact-xrisk-skeptic", "ai-cmp-xrisk-reasons", "ai-pair-xrisk-forecast", "ai-mcq-xrisk-root", "ai-mcq-xpt-gap" ] }, { "id": "ai-l4p3-copyright", "title": "Copyright and Training Data", "order": 2, "blurb": "Keep acquisition, training, outputs, and jurisdiction separate.", "studyGuideAnchor": "copyright-and-training-data", "itemIds": [ "ai-fact-copyright-question", "ai-fact-copyright-split", "ai-cmp-copyright-claim", "ai-pair-acquisition-use", "ai-mcq-copyright-status", "ai-def-fair-use", "ai-cloze-fair-use" ] }, { "id": "ai-l4p4-timelines", "title": "Timelines to AGI", "order": 3, "blurb": "Read a forecast as a conditional distribution, not a countdown.", "studyGuideAnchor": "timelines-to-agi", "itemIds": [ "ai-fact-agi-definition", "ai-fact-agi-survey", "ai-cmp-agi-date", "ai-tf-agi-countdown", "ai-mcq-timeline-variable", "ai-cloze-forecast", "ai-pair-median-forecast" ] } ], "objectives": [ { "id": "ai-obj-disagreements", "statement": "Explain why credible estimates differ on jobs, catastrophic risk, copyright, and AGI timelines.", "demonstrationIds": [ "ai-demo-disagreements" ] } ], "demonstrations": [ { "id": "ai-demo-disagreements", "label": "Reading disputed AI claims", "itemIds": [ "ai-mcq-job-unit", "ai-mcq-estimate-conflict", "ai-mcq-xrisk-root", "ai-mcq-xpt-gap", "ai-mcq-copyright-status", "ai-mcq-timeline-variable" ], "requiredCorrect": 5 } ], "objectiveTests": [ { "id": "ai-test-disagreements", "objectiveId": "ai-obj-disagreements", "title": "The Open Debates", "mcqIds": [ "ai-mcq-job-unit", "ai-mcq-estimate-conflict", "ai-mcq-xrisk-root", "ai-mcq-xpt-gap", "ai-mcq-copyright-status", "ai-mcq-timeline-variable" ] } ] } ] }