VISUAL01
A picture should answer the next question.
See the method before you read the detail.
This guide turns the Playbook's core loop into a route you can inspect: understand, frame, act, inspect, repair, and transfer.
The boards are orientation aids. The localized text and the source page remain the baseline explanation; a picture does not prove that a model acted or that a learner mastered a method.
02
A guided visual route
Know what each picture is for.
Follow the route from your purpose to a checked decision. Select a step to see what it explains, what to ask next, and where to continue.
Next question
Jump to this sectionRead the visual route as text
The same order remains available without JavaScript or pointer interaction.
- 01 · Choose a purposeStart with a safe first task, an uncertain result, a claim to verify, or a method to reuse.Next question: Which small outcome do I need?Jump to this section ↘
- 02 · See the orderFinish the foundation before adding optional platform-specific layers.Next question: Am I still on the foundation route?Jump to this section ↘
- 03 · Separate the termsKeep token, context, prompt, response, and tool authority in their own lanes.Next question: What can this term actually establish?Jump to this section ↘
- 04 · Check the action boundarySeparate a model proposal, host authority, tool execution, human read-back, and evidence.Next question: Which layer can I actually observe?Jump to this section ↘
- 05 · Classify the claimSeparate text, run, external-effect, and unknown claims before choosing the smallest useful check.Next question: Which record could change this claim?Jump to this section ↘
- 06 · Follow the work loopMove through understand, frame, act, inspect, repair, and transfer.Next question: What evidence follows this action?Jump to this section ↘
- 07 · Decide from the recordName the question, locate the source, observe the result, decide within scope, and stop when proof is missing.Next question: What does the record actually support?Jump to this section ↘
- 08 · Browse one decisionOpen one board and its matching lesson; the gallery is not a second catalogue.Next question: Which decision does this board clarify?Jump to this section ↘
- 09 · Read the board correctlyUse the thesis and text fallback, try one bounded move, and keep the boundary visible.Next question: What does the picture still not prove?Jump to this section ↘
- 10 · Read one page into practiceTurn one page into one checked attempt instead of another random click.Next question: What is the smallest attempt I can keep and inspect?Jump to this section ↘
- 11 · Keep a first-attempt receiptRecord the goal, safe input, actual run, focused check, result, limit, and next step.Next question: What can another person inspect?Jump to this section ↘
03
Start from your question
Choose a reason, not a random page.
Pick the situation that brought you here. The map suggests one bounded next step and one visual explanation instead of sending you into the full catalogue.
Next question
Open the route visual
Read the four entry points as text
The same choices remain available as a short list when the map cannot run.
- 01 · Start safelyUse a fictional offline task and keep the first response as your baseline.Open this route ↗
- 02 · Check an uncertain resultPreserve the output, compare it with evidence, and keep the claim inside the record.Open this route ↗
- 03 · Verify a claimBound the claim, name the source, run the smallest check, and record the limit.Open this route ↗
- 04 · Reuse the methodRepeat the method on a new task; one successful attempt is not mastery.Open this route ↗
04
Learning journey map
See the whole Playbook before you choose a track.
One map answers the first practical question: what should I do now, and what comes after the first checked result?
Next move
Open the journey board
Read the four stages as text
Use this ordered list without the interactive map. Each stage names the work and the evidence boundary that follows it.
- 01 · Foundation CoreUnderstand the model, make one request, recognize visible failures, repair, and try a new task.Open this stage ↗
- 02 · First bounded taskName the result, context, allowed help, limits, check, and stop condition.Open this stage ↗
- 03 · Evidence loopCompare what changed with a source, test, log, or acceptance rule; stop when proof is missing.Open this stage ↗
- 04 · Optional tracksChoose Codex, tools, Skills, Agents, research, engineering, or team practice only when the next layer is useful.Open this stage ↗
The map shows order, not mastery. Keep the artifact, evidence, limit, and next question.
05
Teaching boards
Pictures that explain one decision at a time.
Each board has a short explanation, a localized alternative text, and a route into the matching lesson. Open the image when you want the printable version; use the text beside it when you want the exact boundary.
Open selected board
Next question
Read board choices as text
The same board choices remain available without the interactive selection.
- 01 · Read one page as a four-step routeUse the page map to name the next safe action before you open the full catalogue.Open the foundation guide ↗
- 02 · A prompt is a small contractState the goal, context, allowed help, limits, check, and stop condition before asking.Open the task protocol ↗
- 03 · A model proposal is not a tool receiptSeparate proposed text from host authority, tool execution, and the result you can inspect.Open action boundaries ↗
- 04 · A claim can travel only as far as its evidenceKeep a claim inside the source, test, diff, log, or acceptance rule that supports it.Open verification and recovery ↗
- 05 · Failed interaction recoveryPreserve the last known state, classify the failure, and choose a bounded recovery or stop.Open the agent loop lesson ↗
- 06 · From understanding to transferRepeat the method on a new task and keep the evidence that shows what transferred.Open the transfer lesson ↗
A selected picture is an orientation aid. The source page and your own record remain the evidence boundary.
Open more teaching boards
These additional boards cover source checks, handoffs, stopping decisions, and the reading loop. They are optional references, not a second course route.
This is a focused selection from the project-authored teaching boards. It is not a quality score, benchmark, or proof of learning.
06
Capability ladder
Choose the next capability, not the biggest tool.
The ladder turns the learning path into seven observable practices. Select a level to see what it adds, which lesson to open next, and what the level still does not prove.
Observer
Understand the model before you act.
Explain what an LLM, context, prompt, product, tool, Agent, and Skill can establish before choosing a platform.
Level 0 · 1 chapter · 1 lab
Next practiceOpen the first bounded task.
Open the capability ladder
Read the seven levels as text
The same progression remains available without the interactive map. Each level names a practice and a bounded next route.
- 00 · ObserverNotice what happened and keep the visible result separate from what you infer.Open the next lesson ↗
- 01 · Safe userMake one small, reversible request with fictional input, a clear stop, and a saved first response.Open the next lesson ↗
- 02 · Task designerWrite the result, limits, check, and stop so another person can inspect the task contract.Open the next lesson ↗
- 03 · Workflow designerSequence actions around evidence and separate proposal, authority, execution, observation, and recovery.Open the next lesson ↗
- 04 · Capability builderGive a Skill a trigger, boundary, output, evidence, and rollback rule.Open the next lesson ↗
- 05 · Evidence reviewerJudge claims against source, scope, version, access, and visible unknowns.Open the next lesson ↗
- 06 · Team coachMake the method repeatable for others, review the receipt, and stop unsafe escalation.Open the next lesson ↗
A level is a practice target, not a status claim. The project has not measured mastery or transfer.
07
Concept mind map
Separate the six terms before you trust the answer.
Token, context, context window, prompt, response, and tool or Agent are related, but they are not interchangeable. Select one term to see what it explains and what it still cannot prove.
Next check
Open the full concept board
Read the six terms as text
The same six terms remain available as an ordered list. Use the list when the map or image is unavailable.
- 01 · TokenA unit of text processed by the model; it is not a fact or a permission.Next check: What material is actually present?Open the foundation lesson ↗
- 02 · ContextMaterial available in this turn, such as a prompt, supplied text, or tool data.Next check: What is missing and still unknown?Open the foundation lesson ↗
- 03 · Context windowThe amount of material that fits in one turn; the exact limit depends on the product.Next check: Which limit is documented for this surface?Open the foundation lesson ↗
- 04 · PromptThe goal, context, constraints, and requested answer shape; it frames work but grants no access.Next check: What is allowed before the request starts?Open the task protocol ↗
- 05 · ResponseText proposed for inspection; fluent wording can still make an error sound certain.Next check: What record can I compare with the response?Open verification and recovery ↗
- 06 · Tool / AgentA product may add reading, action, or coordination; runtime permission and evidence decide what happened.Next check: What action receipt or log exists?Open action boundaries ↗
Response is not action, and action is not a verified result. Compare the result with a source, diff, test, log, or acceptance rule before claiming completion.
08
Action boundary map
A proposal is not proof that an action happened.
Follow one request through the layers that can be checked: a model proposes words, the host decides authority, a tool may run, a person reads the result, and evidence sets the claim limit.
Next check
Open the action-boundary lessonRead the five boundary layers as text
Use the ordered list if the interaction is unavailable. Each layer names what can be observed and what it still cannot prove.
- 01 · Model proposesWords can name an action, but a plan or tool call is still a proposal.Next check: What did the model actually propose?Open the action-boundary lesson ↗
- 02 · Host decides authorityThe host and a person set the target, permission, and stop boundary.Next check: What authority was granted?Open the action-boundary lesson ↗
- 03 · Tool executesLook for the actual event and returned data; do not infer a run from fluent text.Next check: What event or receipt is visible?Open the action-boundary lesson ↗
- 04 · Person reads backCompare the result with a diff, log, source, test, or acceptance rule.Next check: What did the read-back confirm?Open the action-boundary lesson ↗
- 05 · Evidence limits the claimKeep the claim inside the receipt; if proof is missing, stop and keep it small.Next check: What is the strongest label the record permits?Open the action-boundary lesson ↗
Open the full action-boundary board
If a receipt is missing, keep the claim small. A model's wording cannot establish host permission, tool execution, or correctness.
09
Response claim triage
Classify the claim before you choose a check.
A response can be text, a run claim, an external effect, or an unknown dressed up as certainty. Choose the branch first; it tells you which record can change your claim.
Smallest next check
Open the matching lessonRead the four claim types as text
The same four branches remain available without the interactive map. Start with the smallest record that could change what you say.
- 01 · Text onlyCheck the source, scope, date, and freshness. A fluent answer is not a citation.Smallest next check: Which passage supports the claim?Open the matching lesson ↗
- 02 · Tool or run claimLook for returned data, a diff, a log, or an execution receipt.Smallest next check: What event is visible?Open the matching lesson ↗
- 03 · External side effectConfirm the target, permission, receipt, and recovery path before the effect.Smallest next check: Who authorized this target?Open the matching lesson ↗
- 04 · Unknown or overconfidentUse candidate or unknown, stop, or ask for one smaller check.Smallest next check: What is still unobserved?Open the matching lesson ↗
Open the full claim triage board
Classification is a starting decision, not proof. Keep the record, label uncertainty, and do not infer execution from fluent wording.
10
Dynamic route map
Choose a stage, then ask one question.
Select a node to see the action, the evidence boundary, and the next safe question. The ordered list below keeps the route available without interaction.
Next question
Open this part of the routeRead the six stages as text
Use this ordered route without the interactive map. Each stage names an action and the evidence boundary that follows it.
- 01 · UnderstandSeparate a model proposal from what the available evidence can establish.Open this part of the route ↗
- 02 · FrameState the goal, context, allowed help, limits, check, and stop condition.Open this part of the route ↗
- 03 · ActAllow only the smallest reversible action and pause before an external effect.Open this part of the route ↗
- 04 · InspectCompare the output or diff with a source, test, log, or acceptance rule.Open this part of the route ↗
- 05 · RepairName one mismatch, preserve the failed receipt, and change one condition before retrying.Open this part of the route ↗
- 06 · TransferRepeat the method on an unseen task; one successful attempt is not mastery.Open this part of the route ↗
11
Evidence decision map
Let the record decide how far the claim can go.
This is the practical bridge between a plausible answer and a defensible conclusion: name the question, locate the source, observe what happened, choose a bounded decision, then stop when the next proof is missing.
Next check
Open the evidence routeRead the evidence path as text
The same five steps remain available without the interactive map. Each step says what you can claim and what remains outside the record.
- 01 · QuestionName the decision before you search. A broad topic is not yet a checkable question.Next check: What decision would this evidence change?Open the evidence route ↗
- 02 · SourceName the owner, location, date, version, and boundary of the source before relying on it.Next check: Can I locate the exact passage?Open the evidence route ↗
- 03 · ObserveRead the page, diff, test, log, or response. Do not infer an event you did not observe.Next check: What happened in the record, not in the story?Open the evidence route ↗
- 04 · DecideKeep the claim supported when the check matches; otherwise mark it candidate, unknown, or out of scope.Next check: What is the strongest label the record permits?Open the evidence route ↗
- 05 · StopWhen the next proof is missing, stop and record the smallest safe next check. Preserve the unknown.Next check: What remains unknown, and who can authorize the next step?Open the evidence route ↗
Open the full evidence board
The map is a reasoning aid. It does not prove that a source is correct, a tool ran, or a learner mastered the method.
12
Page-to-practice loop
Turn one page into one checked attempt.
Use the page in order: name the problem, find the concept, try one bounded action, inspect the record, state the boundary, and transfer the method to a changed task.
Next question
Open the matching lessonRead the page-to-practice loop as text
The same six steps remain available as an ordered list. Follow the links only after you can name the next question.
- 01 · Name the problemWrite the concrete question before you browse or ask for a solution.Next question: What would count as a small, observable result?Open the matching lesson ↗
- 02 · Find the conceptLocate the distinction or rule that explains the problem; do not skip straight to a tool.Next question: Which idea changes what I will do?Open the matching lesson ↗
- 03 · Try one bounded moveMake the smallest reversible attempt with a clear input, permission, and stop condition.Next question: What exactly will I change or leave untouched?Open the matching lesson ↗
- 04 · Inspect the recordRead the output, diff, source, test, or log; separate what you saw from what you inferred.Next question: What does the record support?Open the matching lesson ↗
- 05 · State the boundaryName what remains unknown, unauthorized, unobserved, or outside the task.Next question: What is the smallest safe next check?Open the matching lesson ↗
- 06 · Try a changed taskVary one condition and repeat the method without claiming mastery from one success.Next question: What changed, and did the method still hold?Open the matching lesson ↗
Open the page-to-practice board
A completed reading loop is not proof of mastery. Keep the attempt, result, and unknowns visible.
13
First-attempt receipt
Do the small attempt. Keep the receipt.
This five-step map turns a safe first try into a record another person can inspect. Select a step to see what to keep, what to ask next, and where the matching lesson begins.
Next question
Open the matching lessonRead the first-attempt receipt as text
The same five steps remain available as an ordered list. Keep the record smaller than the original task, and stop when the boundary is unclear.
- 01 · Name one checkable resultChoose a small outcome that can be observed without expanding the task.Next question: What would count as a result?Open the matching lesson ↗
- 02 · Keep only safe, relevant inputUse a fixture or non-sensitive example, and mark what is missing instead of guessing.Next question: What input is allowed?Open the task protocol ↗
- 03 · Save what actually happenedKeep the response, diff, exit status, log, or screenshot—not only the plan.Next question: What did I observe?Open verification and recovery ↗
- 04 · Compare one important detailUse a source, diff, test, log, or acceptance rule for one focused check.Next question: Which record could change the claim?Open verification and recovery ↗
- 05 · Write result, limit, and unknownRecord the result and the smallest safe next step, or stop without filling the gap.Next question: What can another person inspect?Open action boundaries ↗
Open the first-attempt receipt board
A receipt shows what was recorded for one bounded attempt. It does not prove that the task was correct, transferable, or mastered.
14
Read the picture correctly
Use a board without overtrusting it.
The board gives you a shape to remember. Your own small attempt and record decide what you may claim.
- 01Start with the thesis
Name the relationship or decision the picture is meant to clarify.
- 02Open the text fallback
Read the labels and next question in ordinary text; do not rely on color or layout alone.
- 03Try one bounded move
Use the linked lesson to make one small, reversible attempt with a clear stop condition.
- 04Keep the boundary visible
Save the output or diff, compare it with evidence, and state what the picture still cannot establish.
Open full-size visual
15
Evidence maturity
Do not skip the evidence.
A complete page is not the same as a practiced, transferred, independently reviewed method. Select the strongest stage you can honestly support.
Supports
Does not support
Evidence needed to move forward
Open the matching lessonRead the evidence stages as text
The same five stages remain available without the interactive map. Use the strongest label the record permits, not the label you hope to reach.
- 01 · DesignedThe teaching contract is written.Next: render the route and make it reachable.Open the matching lesson ↗
- 02 · RenderedThe page and route work for a reader.Next: observe a learner running the task.Open the matching lesson ↗
- 03 · PracticedA learner runs the task and keeps the result.Next: repeat it on a changed task.Open the matching lesson ↗
- 04 · TransferredThe method survives a changed task.Next: invite an independent rerun.Open the matching lesson ↗
- 05 · Independently reviewedAnother person reruns and checks it under a named scope.Next: keep the reviewed boundary explicit.Open the matching lesson ↗
Open the evidence maturity ladder
This ladder is a disclosure aid, not a score. The project currently has substantial structure and static checks, while learner runs, transfer evidence, and independent review remain separate questions.