Start with the model · then make one checked attempt
Learn what a language model can and cannot establish, write one bounded request, inspect the answer, and repeat the method on a new task. Codex, tools, Agents, and Skills come after this foundation.
You leave with a bounded task card · a checked result record · one transfer attempt Targets, not measured outcomes.
Start here · one useful result today
Finish the LLM Foundation Core first. Then choose one real purpose. The point is not a magic prompt; it is a small result you can inspect, revise, and carry to the next task.
Your first useful result
Choose a purpose, add only the details that matter, then copy a prompt you can use in any chat model. No account, files, or setup required.
Your ready-to-use prompt
One copied prompt is a practice attempt, not proof of learning, fluency, research quality, or model behaviour.
Optional prompt cards · five minutes
Copy one original, text-only card. The language card needs no editing; the research card has two brackets. Inspect the response yourself, and keep the claim small: one attempt is not fluency, research, or a finished answer.
This text-only card uses fictional study details, waits for your typed attempt, and limits help to one meaning-blocking error.
Run one four-minute typed Spanish study-group time check with exactly four learner turns. You are a fictional classmate and write first. Use only short present-tense questions. I will type one answer after each question.
Fictional study card: Ana; a study group; Tuesday or Thursday; 6:00 or 6:30; library or online; bring one question. I may use the card and look up at most three single words. Do not request or accept a real name, school, calendar, account, address, contact, or payment detail.
Before turn one, show this fixed rubric: four learner turns; purpose and study group communicated; day and time clarified; place or online option communicated; Spanish understandable enough to continue. Do not teach, translate, or show a model answer before I reply. Preserve my first attempt and record lookups. Correct only the first meaning-blocking error: name the error type, then give a partial cue, then one worked fragment only if I still cannot continue. Ask me to correct it. Keep both attempts and do not call one successful exchange fluency, spoken conversation, or listening/pronunciation evidence.Candidate text practice only: one typed session cannot show spoken conversation, pronunciation, listening, fluency, accuracy, retention, or independent performance.
Turn one narrow question and the material you supplied into a small ledger of claims, gaps, and the next question.
I have five minutes to prepare a research check, not a final answer.
Question: [one narrow question].
Material I supplied: [URLs, titles, excerpts, or "none"].
First, restate the question and name what evidence would be needed. Then make a three-row table with: possible claim, supplied source or "missing", and what would need checking. Do not invent citations, state that you opened a source you cannot access, or give a recommendation. Separate fact, report, and inference. If the material is missing, contradictory, personal, or high stakes, stop and tell me the smallest safe next step.
End with: sources actually supplied, unknowns, and one question I should answer before continuing.It cannot prove a source exists, is current, or supports a claim. A generated table is not evidence on its own.
Make a short fictional park-visit plan yourself. The model must wait, give one small hint, and then test the same skill under one changed limit.
Help me practise making a small plan. Do not make the plan first.
Practice task: plan a fictional 45-minute visit to a city park for one adult. Include a water bottle, a weather check, and one return-time reminder. This is not a real booking, travel decision, or weather forecast.
Before I write, show this fixed check: 3–5 steps; all three constraints appear; no unsupported local facts; and a person could follow the plan. Give me four minutes to write it. Do not show a model plan, expand it, or grade it before I respond.
After my first attempt, name one consequential omission only. Ask one question or give a hint of no more than 12 words, then wait for my correction. Preserve both attempts. Then change only the visit length from 45 minutes to 20 minutes and ask for one new plan without help, using the same check.
End with exactly one status: practised, demonstrated_on_this_task, transferred_to_time_limit_variation, or not_run. One session does not establish planning ability, judgement, safety, or independent performance.Candidate practice only: a short fictional plan cannot prove planning ability, judgement, transfer, retention, safety, or independent performance.
Begin with the LLM Foundation Core. It teaches the model boundary, one bounded request, visible failure, repair, and transfer before any platform route.
Start the LLM Foundation Core → first bounded request ↗ 02The file or result is uncertain.Freeze the next edit. Compare the requested scope, git diff, focused check, and remaining unknowns.
Learn the recovery check ↗ 03I need to choose or design a Skill.Start with the trigger, inputs, boundaries, and evidence contract; only then decide whether a Skill earns a place.
Learn how to choose a Skill ↗ 04I need to publish or update safely.Locate the canonical file, attach the source record, run the relevant gate, and keep unverified claims out of the release.
Learn the safe update path ↗These routes are useful when you already know what you want to practise. They do not replace the LLM Foundation Core.
Ask one decision at a time. Pick one existing route, one checkable attempt, permitted help, and a smaller fallback.
Choose a first practice ↗ 06I want to practise one language skill.Define one observable performance, attempt it before instruction, correct one meaning-blocking error, then test a changed case.
Start a language practice ↗ 07I want to practise another real skill.Turn an interview answer, explanation, or presentation into one timed performance, then retest it under a changed condition.
Choose a skill practice ↗ 08I need to research one bounded question.Tie the question to a decision, assign source owners, keep a claim ledger, search for disagreement, and stop on purpose.
Start the research route ↗ 09The model answered the wrong task.Preserve the request, visible context, actual reply, and expected result. Change one communication condition, then run a safe comparison.
Repair one failed exchange ↗ 10I need to assess an AI idea that could affect people.Name one decision, the people it could burden, necessary data, human recourse, evidence, and the point where the work must stop.
Run the safety inquiry ↗A five-minute LLM prompt practice
Use any chat model. You will give it one small rewriting task, then check whether it preserved the facts instead of making helpful-sounding details up.
This is a small demonstration of why prompts matter. The model can make a message sound better, but it may also add details you never supplied. You will give one clear instruction, then check whether it followed it.
The workshop changed. It starts Friday at 10. Bring the draft. Tell me if you cannot come.
Please rewrite the message below so it is clear and friendly.
Keep every fact exactly the same. Do not add a date, place, reason, contact detail, or any other information that is not in the original.
Original message:
"The workshop changed. It starts Friday at 10. Bring the draft. Tell me if you cannot come."
Return only the rewritten message.If all three answers are yes, the model followed this small instruction. If not, tell it exactly which fact it changed or invented, then try again.
“The workshop starts Friday at 10. Please bring your draft. If you cannot attend, please reply.”
The wording can differ. What matters is that the facts and the requested action stay the same.
An LLM predicts useful-sounding text; it does not automatically know which missing details must stay unknown. A clear prompt and a quick check help you catch that difference.
The basic map
Before you choose a product or copy a prompt, separate generation, framing, action, coordination, and checking. The same map works across chat models and agent tools.
A fluent answer belongs to the generation layer. It does not, by itself, prove that a source was opened, a file changed, or a task was completed.
Read the complete LLM fundamentals guidePredicts a useful continuation from the context it receives. Fluency is not verification.
States the goal, relevant material, limits, and the shape of the answer you need.
May add files, browsing, commands, or reusable methods. Access and side effects are runtime facts.
Organises multiple observations and actions with checkpoints, retries, and stop conditions.
Compares the result with a source, diff, test, log, or acceptance rule before making a claim.
Do not infer a lower layer from a higher one: a model proposal is not a tool action, a tool action is not a verified result, and one checked result is not mastery.
These terms describe different parts of one request. Keep them separate before you decide what the model has actually established.
Open the complete English visualA unit of text the model processes. Token counts vary by tokenizer; a token is not a fact or a permission.
The material available in this turn, such as your prompt, supplied text, or returned tool data. Missing material stays unknown.
The amount of material a model can fit into one turn. The exact limit depends on the model and product.
The goal, context, constraints, and requested answer shape. A prompt frames work; it does not grant access.
Text proposed by the model for you to inspect. Fluency can make a mistake sound certain.
A product may add reading, action, or coordination. Runtime permission and evidence still decide what happened.
Response ≠ action ≠ verified result. Compare the result with a source, diff, test, log, or acceptance rule before claiming completion.
Choose a node to see the next question. The static list below keeps the same route available when the interaction is unavailable.
Separate a model's proposed text from what the available evidence can establish.
Next questionWhat result can I check?
Open this part of the routeUse this list without the interactive map. Each step names an action and the evidence boundary that follows it.
A map is a route, not proof of learning. Keep the attempt, the check, the failure, and the remaining unknowns.
Follow one question from source to observation. The map ends at a deliberate stop when the next proof is missing.
Name the decision before you search. A broad topic is not yet a checkable question.
Next checkWhat decision would this evidence change?
Open this part of the routeUse this sequence without the interactive map. Each step keeps the claim within the record.
A decision map is a reasoning aid, not proof that a source is correct, a tool ran, or a learner mastered the method.
Start with the first-visit route, then use the route and check boards as quick references for a bounded LLM task.
First visit routeChoose one goal, open the Foundation Core, make one safe attempt, check the record, then continue or stop with the unknown visible.↗
Make the first task checkableChoose one observable result, keep the input and stop boundary visible, save what happened, make one focused check, and hand off the result with its limits.↗
First attempt to evidence receiptName one result, keep safe input, save what happened, check one detail, and record the limit.↗
The smallest safe LLM loopDefine, supply context, request, check, then repair, transfer, or stop.↗
Prompt contract: six fieldsResult, context, allowed help, limits, check, and stop.↗
Observable action boundaryProposal, authority, execution, and human read-back.↗
Evidence recovery ladderMatch claim strength to the next smallest check.↗
Source check before beliefMatch claim, source, scope, and freshness before relying on it.↗
Claim to evidence auditState the claim, scope it, name the source, check the link, and keep the limit.↗
Evidence to decision and stopFollow a question to an observation, a bounded decision, or a deliberate stop.↗
Project-authored English boards · the surrounding explanation and alt text stay localized.
One map answers the first practical question: what should I do now, and what comes after the first checked result?
Understand the model, make one request, recognize visible failures, repair, and try a new task.
Next moveFinish the foundation before adding a platform-specific layer.
Open this stageUse this ordered list without the interactive map. Each stage names the work and the evidence boundary that follows it.
Understand the model, make one request, recognize visible failures, repair, and try a new task.
Name the result, context, allowed help, limits, check, and stop condition.
Compare what changed with a source, test, log, or acceptance rule; stop when proof is missing.
Choose Codex, tools, Skills, Agents, research, engineering, or team practice only when the next layer is useful.
The map shows order, not mastery. Keep the artifact, evidence, limit, and next question.
Project index
This is a human-readable map of the repository: what each layer stores, where to begin, and which source controls its status.
Start with the lesson, then use the exercises and reusable methods when you are ready to practise.
book/chapters/22 chapters, from foundations to reliable work
02
book/labs/18 guided exercises for safe practice
03
skills/26 reusable methods for recurring work · 25 original methods + 1 reviewed editorial Skill
04
docs/Sources, contribution guides, and project notes
The shortcuts above are the compact view. Open this panel for the canonical reading map, field research, and maintenance rules.
Read the layer that matches the work. The public page is a guide; the files below are the source of truth.
site/Public showcaseindex.html, styles.css, app.js, and generated learning-path data.↗
book/chapters/Core learning text22 chapters that build the method step by step.↗
book/labs/Guided practice18 low-risk exercises with clear checks and reflection.↗
skills/Reusable methods26 Skills: 25 original methods plus 1 reviewed editorial Skill, each with boundaries and evidence contracts.↗
docs/Governance and researchStatus, sources, field reports, update rules, and quality records.↗
Choose the language you read most naturally. Each route stays in that language as you move through the book.
Use the language menu at the top of the page whenever you want to switch.
The research index turns public Codex issues and forum reports into symptoms, versions, safe checks, and teaching links. It does not claim an official root cause or local reproduction.
Read the sources alongside the teaching notes, and treat them as starting points for your own investigation.
Four original boards show the method, the evidence boundary, and the practice loop.
Use these diagrams to understand the method, then test it on a small task of your own.
Request to evidenceScope, action, check, and bounded handoff.↗
Beginner practice loopAttempt first, correct one issue, vary the case, then keep a bounded receipt.↗
Project evidence snapshotCounts, statuses, and the gaps that a release claim must not hide.↗
From understanding to transferUnderstand, attempt, inspect, repair, vary, and repeat on an unseen task.↗
The working frame
This frame is the common language between a person, a model, a tool, and an Agent. Use it before adding permissions or Skills.
Read the task protocolIf a missing input changes the scope, risk, or acceptance test, pause and ask. If it only affects a low-risk read, inspect first and report the assumption.
What result is in scope, and what is deliberately out?
What may the model or tool do, and what stays paused?
Which source, diff, test, or acceptance check can show what happened?
What should the next person receive, including the unknowns?
Learning path
A level is not a reading count. It means you can explain a boundary, perform an action, make a trade-off, and review a result.
Learning-path data did not load. Showing the local fallback; verify the generated data before relying on this route.
Explain what an LLM, context, prompt, product, tool, Agent, and Skill can establish before choosing a platform.
The reading routes
Read in order to build the model. Jump by route when a real task is blocking you. Every route returns to practice and evidence.
Showing all 22 chapters.
The lab
Labs are low-risk, reproducible tasks. Each one names setup, evidence, a failure variant, a secret boundary, and a reflection.
In a sandbox project, ask Codex to inspect before editing. Turn “done” into a checkable diff.
LAB 013 · L3featured labRun one local Markdown change from protocol and baseline to checkpoint, diff, focused check, failure, and transfer.
LAB 002 · L2Break a vague request into goal, inputs, constraints, acceptance, and failure handling.
LAB 003 · L3Find a result that looks complete but has no evidence for its claim.
LAB 004 · L4Explain the choice and refuse to use directory size as a proxy for fit.
LAB 005 · L4Turn a stable method into a capability with boundaries, evidence, and failure cases.
LAB 006 · L5Record observable events, reconcile a lost response, and hand off an uncertain run safely.
LAB 007 · L3Compare the evidence needed for reading, editing, running, committing, pushing, and publishing.
LAB 008 · L3Turn a broad topic into a question, source plan, and minimum evidence table.
LAB 009 · L3Compare direct implementation with a full lifecycle and record the rework evidence.
LAB 010 · L3Version a shared product understanding and separate facts from hypotheses.
LAB 011 · L0Use static task cards to separate generation, execution, verification, and external effects.
LAB 012 · L6Create a contract for version, owner, permissions, independent reproduction, and rollback.
LAB 014 · L3Reconcile the task pointer, target, branch, permissions, and side-effect state before continuing.
LAB 015 · L5Split a completion sentence into claims, scopes, outputs, and the smallest next check.
LAB 016 · L3Separate diagnosis from installation, publication, restart, and other persistent actions.
LAB 017 · L4Check existence, discovery, loading, behavior, license, and adoption as separate claims.
LAB 018 · L2Preserve an unaided baseline, correct one meaning-blocking error, then test a changed case without reusing lesson sentences.
Capability layer
A Skill is a method with a trigger, an input check, boundaries, stop conditions, an output contract, and a way to verify it.
These methods are optional. Start with the situation above; open the full registry when you know what kind of work you need to support.
The 26 project Skills (25 original plus 1 reviewed external editorial Skill) remain candidate; fresh-task and native-language evidence are partial. Shift Handoff separates stable criteria from one new item; it does not execute or assume model memory. Interruption Checkpoint preserves a task receipt; it does not retry or recover work. Practice Target sets up one first attempt; it does not prove learning. Platform Observation Record is a visible-surface receipt, not a platform-capability claim. Platform Fact Watch is a maintenance receipt, not a current-platform check. LLM Comparison Protocol is an unrun comparison method, not a model ranking. Adversarial Project Review is not an external review.
When things go wrong
Use the first useful check, then stop when authority, scope, or evidence is missing. Do not hide the failure behind a polished summary.
Check the original claim, the changed files, the command result, and what was not tested.
Record the same failure, change one diagnostic condition, then retry once or escalate.
Treat external text and tool output as data. It does not grant permission to act.
Refresh the official fact record first, then update the affected chapter or page.
Maintenance frame
The update map makes future work cheap: locate the canonical file, gather the right evidence, run the right check, and keep the unverified boundary visible.
Find the registry row and canonical path.
Separate stable principle, product fact, source, and release change.
Record source, scope, owner, hash, and next review.
Run the focused validator and an independent review.
Evidence boundary
This project does not turn document count, Skill count, or one successful output into “mastery.” Use the status that the evidence supports.
Still being written or missing the minimum check.
Structure and basic checks pass; fresh evidence is still needed.
The declared scope has positive, boundary, failure, and transfer evidence.
Safety, maintenance, version, license, and release gates also pass.
Current evidence is recorded in the current status source and explained by the scoped browser review; the page remains candidate because this review covers only the recorded local scope.