mirror of
https://github.com/byoungd/up.git
synced 2026-10-02 08:04:40 +08:00
Harden AI task and learning evidence workflows
This commit is contained in:
Leap 离谱
@@ -8,6 +8,8 @@ All notable project-level changes are documented here. Content pages retain thei
|
||||
|
||||
#### Changed
|
||||
|
||||
- Expanded the bilingual AI Task Brief and AI Learning Log into a governed workflow covering evaluation sets, human gates, data retention, cost, model scope, source verification, failure rollback, handover, independent production, speed debt, and delayed transfer.
|
||||
- Connected the AI work papers through the glossary and regression coverage so a model response cannot masquerade as a verified deliverable or independent learning.
|
||||
- Rebuilt the bilingual Learning Principles chapter and English Diagnostic around real deliveries, first-version conditions, error diagnosis, targeted repair, feedback uptake, performance-adjusted spacing, recovery and capacity modes, cross-session state, AI boundaries, and one-condition transfer.
|
||||
- Renamed the bilingual diagnostic's vocabulary entry as an evidence card and connected the general protocol through the Reader's Guide, glossary, search index, attribution register, and regression coverage.
|
||||
- Rebuilt the bilingual Vocabulary chapter and Evidence Card around real-task baselines, five decisions for unknown items, eight dimensions of word knowledge, receptive/productive gaps, coverage-number limits, performance-adjusted spacing, error repair, AI verification, and fourteen-day transfer.
|
||||
|
||||
@@ -67,6 +67,8 @@ function splitSearchSections(file: string, html: string) {
|
||||
"/templates/reading-evidence.md",
|
||||
"/templates/vocabulary-audit.md",
|
||||
"/templates/english-diagnostic.md",
|
||||
"/templates/ai-task-brief.md",
|
||||
"/templates/ai-learning-log.md",
|
||||
"/threads/part-4/family-learning.md",
|
||||
"/threads/part-1/8-job-search-english.md",
|
||||
"/threads/part-1/grammar.md",
|
||||
|
||||
@@ -19,6 +19,7 @@ Find the task you are facing, then follow “definition → evidence → next st
|
||||
| **Real problem** | A situation with affected people and an action to complete, not “I want to improve” | [Learning Principles](../threads/part-1/1-understanding.md) |
|
||||
| **Context** | When, where, and for whom a skill will be used | [CEFR Goals and Self-check](../threads/part-1/0-cefr.md) |
|
||||
| **Acceptance standard** | An outcome another person can observe, check, or restate | [AI Task Brief](../templates/ai-task-brief.md) |
|
||||
| **Evaluation set** | Normal, boundary, missing/conflicting, and realistic redacted samples used to check output instead of showing only model-friendly examples | [AI Task Brief](../templates/ai-task-brief.md) |
|
||||
| **Main question** | One answerable question tracked for a week | [90-Day Action Plan](../threads/part-5/90-day-plan.md) |
|
||||
|
||||
## Evidence and Judgment
|
||||
@@ -90,7 +91,9 @@ Find the task you are facing, then follow “definition → evidence → next st
|
||||
| **Interaction repair** | When hearing, scope, or knowledge is incomplete, request repetition, confirm, ask for thinking time, and summarise agreement | [Speaking](../threads/part-1/5-speaking.md) · [Job-search English](../threads/part-1/8-job-search-english.md) |
|
||||
| **Artifact** | An output others can read, use, question, or improve; not necessarily a product | [Artifacts](../threads/part-3/4-artifacts-and-delivery.md) |
|
||||
| **Delivery** | Completing work under a real audience, user, or constraint and accepting the result | [90-Day Action Plan](../threads/part-5/90-day-plan.md) |
|
||||
| **Human gate** | AI may assist, but a person confirms facts, permissions, privacy, cost, and final judgment | [AI Project Development](../threads/part-3/2-ai-development-and-resource-layer.md) |
|
||||
| **Human gate** | AI may assist, but named people confirm source, fact, permission, privacy, quality, cost, and final ownership before release or execution | [AI Task Brief](../templates/ai-task-brief.md) · [AI Project Development](../threads/part-3/2-ai-development-and-resource-layer.md) |
|
||||
| **Speed debt** | Real cost owed when shorter immediate time creates more rework, verification, communication, or incident risk | [AI Learning Log](../templates/ai-learning-log.md) |
|
||||
| **AI Learning Log** | A private sheet comparing unaided, assisted, delayed-independent, and parallel tasks to record tool contribution and owned ability | [AI Learning Log](../templates/ai-learning-log.md) |
|
||||
| **Rollback** | Pause, degrade, switch, or return to a known usable state after failure | [AI Learning and Project Practice](../threads/part-3/1-ai-learning.md) |
|
||||
|
||||
## Life and Boundaries
|
||||
|
||||
@@ -1,53 +1,131 @@
|
||||
---
|
||||
title: AI Learning Log Template
|
||||
description: Preserve the unaided baseline, guided practice, errors, evidence, independent performance, and transfer so learning can continue and be retested.
|
||||
updated: 2026-08-31
|
||||
title: "AI Learning Log: From Tool Collaboration to Independent Ability"
|
||||
description: Record a real learning task, unaided baseline, tool scope, source checks, active production, delayed retest, transfer, and the next cycle decision.
|
||||
updated: 2026-09-02
|
||||
---
|
||||
|
||||
# AI Learning Log Template
|
||||
# AI Learning Log: From Tool Collaboration to Independent Ability
|
||||
|
||||
Record one inspectable learning slice at a time. An AI-generated summary is not a source of truth; original work and primary sources are.
|
||||
Record one inspectable learning slice at a time. An AI-generated summary, score, or explanation is not a source of truth; original material, your first version, actual work, and real feedback are evidence. Do not write “what I discussed with the model”. Write “what I can still complete independently after the tool leaves”.
|
||||
|
||||
## 1. Task and Starting Point
|
||||
|
||||
```markdown
|
||||
# AI Learning Log — YYYY-MM-DD
|
||||
# AI Learning Log - YYYY-MM-DD
|
||||
|
||||
## Task
|
||||
- Real problem:
|
||||
- Acceptance criteria:
|
||||
- Smallest slice:
|
||||
|
||||
## Baseline
|
||||
- Unaided attempt:
|
||||
- Gaps exposed:
|
||||
- Material and sources:
|
||||
|
||||
## AI collaboration
|
||||
- What AI did:
|
||||
- Judgments I kept:
|
||||
- Answers that require verification:
|
||||
|
||||
## Active production
|
||||
- Work completed after closing AI:
|
||||
- Independent time:
|
||||
- Tests, feedback, or score:
|
||||
|
||||
## Three comparisons
|
||||
| Sample | Conditions | Task/quality | Time/rework | Confidence | Evidence location |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| Unaided baseline | | | | | |
|
||||
| Assisted version | | | | | |
|
||||
| Delayed independent (after 3–7 days) | | | | | |
|
||||
|
||||
## Errors and transfer
|
||||
| Error/gap | Evidence | Possible cause | Parallel-task result |
|
||||
| --- | --- | --- | --- |
|
||||
| | | | |
|
||||
|
||||
## Next step
|
||||
- Smallest next task:
|
||||
- Acceptance criteria:
|
||||
- Required material:
|
||||
- Evidence to preserve:
|
||||
- Handover owner/date:
|
||||
- Stop or rollback condition:
|
||||
Real situation and audience:
|
||||
Main question:
|
||||
Action and acceptance standard:
|
||||
Material, version, and source:
|
||||
Time, device, and environment limits:
|
||||
Permitted dictionary / translation / search / AI / human help:
|
||||
Privacy, copyright, exam, and ownership boundary:
|
||||
```
|
||||
|
||||
## 2. Unaided Baseline
|
||||
|
||||
```markdown
|
||||
Start/end time of independent attempt:
|
||||
Unaided sample location:
|
||||
What I could explain or complete:
|
||||
Errors, unknowns, and judgment blind spots exposed:
|
||||
Confidence (0-2):
|
||||
```
|
||||
|
||||
Do not ask AI to write an answer first and call that the starting point. If the task permits a dictionary or documentation, record the permitted support; the baseline states what you can do independently under those conditions.
|
||||
|
||||
## 3. Agree What AI Does
|
||||
|
||||
```markdown
|
||||
Model / version / date:
|
||||
Mode: diagnosis / follow-up / explanation / counterexample / transcript / candidate / other
|
||||
AI may:
|
||||
AI may not:
|
||||
Minimum material sent:
|
||||
Fields or files withheld:
|
||||
```
|
||||
|
||||
Move one slice per round. Ask the model to identify gaps, assumptions, sources, and uncertainty before requesting any rewrite or candidate.
|
||||
|
||||
## 4. Interaction and Source Verification
|
||||
|
||||
| Turn | My question/action | Model suggestion | Type | Source/uncertainty | My decision |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| | | | explanation / question / candidate / feedback / code / other | | accept / partly / reject / verify |
|
||||
|
||||
```markdown
|
||||
Primary sources I checked:
|
||||
Errors, outdated claims, or invented material I found:
|
||||
Suggestions that changed my understanding and why:
|
||||
Suggestions that changed only surface wording:
|
||||
```
|
||||
|
||||
People confirm sources, facts, permissions, privacy, and final judgment. Model confidence does not raise the evidence level.
|
||||
|
||||
## 5. Active Production After Closing AI
|
||||
|
||||
```markdown
|
||||
Time AI was closed:
|
||||
Work, explanation, code, or decision completed independently:
|
||||
Independent time:
|
||||
Cues still required:
|
||||
Can I explain every critical step? yes / partly / no
|
||||
Test, reader/user feedback, or run result:
|
||||
```
|
||||
|
||||
## 6. Three Comparisons
|
||||
|
||||
| Sample | Conditions | Task/quality | Time | Rework | Confidence | Evidence location |
|
||||
| --- | --- | --- | ---: | ---: | ---: | --- |
|
||||
| Unaided baseline | | | | | | |
|
||||
| AI-assisted version | | | | | | |
|
||||
| Delayed independent (3-7 days) | | | | | | |
|
||||
| Parallel task (changed condition) | | | | | | |
|
||||
|
||||
| Result | Cautious interpretation |
|
||||
| --- | --- |
|
||||
| Assisted and independent versions improve | A removable scaffold may be forming; test more transfer |
|
||||
| Only assisted version improves | Product improved; critical steps may be outsourced |
|
||||
| Time falls while rework rises | Speed created speed debt |
|
||||
| Confidence rises while facts/tests fall | Calibrate judgment before increasing tool access |
|
||||
|
||||
One comparison cannot prove causality, but it records contribution more clearly than “AI was useful”.
|
||||
|
||||
## 7. Errors, Feedback, and Transfer
|
||||
|
||||
| Error/gap | Sample | Type | Change only this next | Parallel-task result |
|
||||
| --- | --- | --- | --- | --- |
|
||||
| | | fact / vocabulary / structure / grammar / strategy / attention / tool dependence | | |
|
||||
|
||||
```markdown
|
||||
Real reader/user retelling or execution result:
|
||||
Feedback accepted, partly accepted, rejected, or deferred and why:
|
||||
Action that survived a changed topic/audience/mode/time:
|
||||
Action that still succeeds only on the original task:
|
||||
```
|
||||
|
||||
## 8. Delayed Retention and Handover
|
||||
|
||||
```markdown
|
||||
Day 1: what I removed and completed independently:
|
||||
Days 3-7: what changed and what happened:
|
||||
Day 14: can I explain and use it after closing the old chat:
|
||||
Day 30: did the real task require less guessing or rework:
|
||||
State-file location:
|
||||
Handover owner, date, and access:
|
||||
Stop or rollback condition:
|
||||
```
|
||||
|
||||
## 9. Next Cycle Decision
|
||||
|
||||
```markdown
|
||||
What the tool genuinely helped:
|
||||
Risk or dependence the tool created:
|
||||
Evidence of ability I now own:
|
||||
Conclusion still unsupported:
|
||||
Continue / downgrade / change variable / pause / seek help:
|
||||
Smallest next task:
|
||||
Next review date:
|
||||
```
|
||||
|
||||
Related entry points: [Learning Principles: Turn Effort into Verifiable Learning](../threads/part-1/1-understanding.md) | [Learning Anything with AI](../threads/part-3/1-ai-learning.md) | [AI Task Brief](ai-task-brief.md) | [Learning State](learning-state.md) | [Evidence Chain](evidence-chain.md)
|
||||
|
||||
@@ -1,48 +1,144 @@
|
||||
---
|
||||
title: AI Task Brief Template
|
||||
description: Define an AI learning or project task with a real situation, data boundary, acceptance criteria, and human responsibility.
|
||||
updated: 2026-08-24
|
||||
title: "AI Task Brief: From Problem to Human Acceptance"
|
||||
description: Before an AI learning or project task, define the real action, data boundary, sources, evaluation set, human ownership, cost, failure handling, and handover.
|
||||
updated: 2026-09-02
|
||||
---
|
||||
|
||||
# AI Task Brief Template
|
||||
# AI Task Brief: From Problem to Human Acceptance
|
||||
|
||||
Copy this into a private project directory. Do not include passwords, identity documents, precise addresses, medical privacy, customer records, or unauthorised third-party information.
|
||||
Copy this into a private project directory before a model call. It is not a prompt collection. It is permission to start: when task, data, sources, acceptance, or ownership remain unclear, do not ask a model for the final result yet.
|
||||
|
||||
Do not include passwords, identity documents, precise addresses, medical privacy, children's data, customer records, unpublished vulnerabilities, or unauthorised third-party material. Redact sensitive material or use an organisation-approved environment.
|
||||
|
||||
## 1. Task and Ownership
|
||||
|
||||
```markdown
|
||||
# AI Task Brief
|
||||
# AI Task Brief - YYYY-MM-DD
|
||||
|
||||
Updated: YYYY-MM-DD
|
||||
|
||||
## Task
|
||||
- Real situation:
|
||||
- User/audience:
|
||||
- Decision or action to complete:
|
||||
- Deadline:
|
||||
|
||||
## Inputs and boundaries
|
||||
- Known facts and sources:
|
||||
- Files allowed:
|
||||
- Data sensitivity: public / internal / confidential / restricted
|
||||
- Material deliberately withheld:
|
||||
- Copyright or permission limits:
|
||||
|
||||
## Output and acceptance
|
||||
- Final deliverable:
|
||||
- Format and length:
|
||||
- Observable definition of done:
|
||||
- Items a person must confirm:
|
||||
|
||||
## Working rules
|
||||
- AI may:
|
||||
- AI may not:
|
||||
- Human reviewer:
|
||||
- How work pauses, rolls back, or notifies on failure:
|
||||
|
||||
## Evidence locations
|
||||
- Unaided baseline:
|
||||
- Primary sources:
|
||||
- Tests/feedback:
|
||||
- Final version:
|
||||
Real situation:
|
||||
User/audience:
|
||||
Decision or action to complete:
|
||||
Why now:
|
||||
Deadline and non-negotiable checkpoints:
|
||||
Final owner and human reviewer:
|
||||
Who is affected if this is delayed or not done:
|
||||
```
|
||||
|
||||
Write acceptance as an action, such as “a user can import a file in ten minutes and see an explainable error report”, not “the experience is good” or “the system is intelligent enough”.
|
||||
Replace “learn AI” or “build an intelligent assistant” with an observable action, such as “a user can import a file in ten minutes and see an explainable report with sources and errors”.
|
||||
|
||||
## 2. Inputs, Sources, and Data Boundary
|
||||
|
||||
| Input/claim | Type | Source, version, location | May reach model? | Verifier | Expiry condition |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| | fact / inference / experience / user data / third-party | | public / redacted / approved / prohibited | | |
|
||||
|
||||
```markdown
|
||||
Files and fields allowed:
|
||||
Material deliberately withheld:
|
||||
Data sensitivity: public / internal / confidential / restricted
|
||||
Collection, transfer, retention, and deletion dates:
|
||||
Required consent obtained:
|
||||
Copyright, licence, citation, and authorship requirements:
|
||||
Minimum data scope the model can see:
|
||||
```
|
||||
|
||||
A source is not automatically true. Each critical claim must return to an original passage, version, data definition, or direct observation. Model links, numbers, and quotations still require human verification.
|
||||
|
||||
## 3. Output and Evaluation Set
|
||||
|
||||
```markdown
|
||||
Final deliverable:
|
||||
Format, length, and audience:
|
||||
Observable definition of done:
|
||||
Hard gates that must pass:
|
||||
Acceptable variation:
|
||||
Content that must not appear:
|
||||
```
|
||||
|
||||
Build a small, realistic evaluation set instead of only model-friendly examples:
|
||||
|
||||
| Sample | Input condition | Expected result | Unacceptable result | Actual result | Evidence location |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| Normal case | | | | | |
|
||||
| Boundary case | | | | | |
|
||||
| Missing/conflicting input | | | | | |
|
||||
| Redacted historical case | | | | | |
|
||||
|
||||
Acceptance should answer whether facts are traceable, the action is complete, errors are visible, permissions are correct, and failure can stop. Fluent, fast, or human-like is not an acceptance standard.
|
||||
|
||||
## 4. AI's Working Scope
|
||||
|
||||
```markdown
|
||||
AI may: ask / classify / propose explanations / give counterexamples / transcribe
|
||||
/ draft / suggest tests
|
||||
AI may not: make final factual decisions / invent sources / approve for the owner
|
||||
/ cross permissions / complete a prohibited exam or application
|
||||
/ publish or execute automatically
|
||||
Mode: diagnosis / assistance / candidate generation / batch processing / other
|
||||
Model, version, region, and call date:
|
||||
Prompt, system instruction, or workflow version location:
|
||||
```
|
||||
|
||||
Ask the model to restate goal, input, limits, unknowns, and acceptance before generation. Version prompts, and never treat one chat window as the project's only record.
|
||||
|
||||
## 5. Human Gates
|
||||
|
||||
| Gate | Who confirms | Passing evidence | If it fails |
|
||||
| --- | --- | --- | --- |
|
||||
| Source gate | | Original link, version, location | Mark unverified; do not circulate |
|
||||
| Fact gate | | Sample checks and data definition | Delete or downgrade claim |
|
||||
| Privacy/permission gate | | Scope, consent, access record | Stop and redact |
|
||||
| Quality gate | | Evaluation set, edge cases, real feedback | Repair, narrow, or reject |
|
||||
| Cost gate | | Tokens, time, human rework, budget | Downgrade or stop |
|
||||
| Ownership gate | | Named approval and disclosure | Do not publish or execute |
|
||||
|
||||
Critical decisions cannot be approved only by the generator, an automated score, or one developer alone. High-risk domains return to current primary sources and qualified professionals.
|
||||
|
||||
## 6. Cost, Retention, and Reversibility
|
||||
|
||||
```markdown
|
||||
Call volume, time, and cost ceiling:
|
||||
Human review and rework budget:
|
||||
Does data enter training, logs, or third-party retention:
|
||||
Acceptable latency and downgrade path:
|
||||
Version that can be withdrawn, rerun, or restored:
|
||||
Release scope and pilot audience:
|
||||
```
|
||||
|
||||
A cheap call that causes a privacy incident, wrong decision, or major rework is not cheap in reality. Record human time, review, failure, reruns, and communication in addition to model fees.
|
||||
|
||||
## 7. Failure, Pause, and Rollback
|
||||
|
||||
| Trigger | Immediate action | Notify | Recovery/rollback location |
|
||||
| --- | --- | --- | --- |
|
||||
| Source or version cannot be found | Stop circulation; return to original | | |
|
||||
| Evaluation hard gate fails | Block release or automatic action | | |
|
||||
| Input crosses permission boundary | Stop upload; revoke access | | |
|
||||
| Cost/latency exceeds ceiling | Downgrade, throttle, or stop | | |
|
||||
| Real user reports harm | Remove, preserve evidence, escalate | | |
|
||||
|
||||
Without a named stop owner, notification path, and known-good version, the task is not ready for a real workflow.
|
||||
|
||||
## 8. Handover and Public Disclosure
|
||||
|
||||
```markdown
|
||||
Current state: not started / experiment / internal pilot / limited release / delivered / stopped
|
||||
Completed work and evidence locations:
|
||||
Open questions and risks:
|
||||
Smallest next task:
|
||||
Handover owner, date, and access:
|
||||
What readers/users need to know about AI involvement:
|
||||
Source, privacy, copyright, and conflict-of-interest note:
|
||||
```
|
||||
|
||||
The next operator should not search chat history to discover what to do. Public work should state which step used a tool, which facts a person confirmed, and which content remains a candidate.
|
||||
|
||||
## 9. Preflight Check
|
||||
|
||||
- [ ] Audience, action, deadline, and owner are explicit.
|
||||
- [ ] Input fields, sensitivity, consent, copyright, and retention are confirmed.
|
||||
- [ ] Critical sources, versions, and expiry conditions are traceable.
|
||||
- [ ] Normal, boundary, conflict, and redacted historical cases are in the evaluation set.
|
||||
- [ ] AI may/may-not scope is explicit.
|
||||
- [ ] Human gates, cost ceiling, stop conditions, and rollback location are named.
|
||||
- [ ] Handover, disclosure, and next review date are set.
|
||||
|
||||
@@ -84,7 +84,7 @@ Without step one, AI turns a wish into an answer. Without step two, you cannot s
|
||||
|
||||
## One-page Task Brief
|
||||
|
||||
Copy this into a private project directory. Redact sensitive material; never send passwords, identity documents, customer records, or third-party private data to a general model.
|
||||
Copy the complete [AI Task Brief](../../templates/ai-task-brief.md) into a private project directory. The block below keeps only the minimum fields. Redact sensitive material; never send passwords, identity documents, customer records, or third-party private data to a general model.
|
||||
|
||||
```markdown
|
||||
# AI Task Brief
|
||||
@@ -155,7 +155,7 @@ For high-risk content, “not yet confirmed” is more professional than a fluen
|
||||
|
||||
## Three Comparisons: Prove What AI Changed
|
||||
|
||||
Keep three samples of the same task:
|
||||
Use the [AI Learning Log](../../templates/ai-learning-log.md) to keep three samples of the same task:
|
||||
|
||||
1. **Unaided baseline**: complete it independently and record time, quality, bottlenecks, and confidence;
|
||||
2. **Assisted version**: let AI do only the agreed work and record prompts, sources, accepted/rejected suggestions, and rework;
|
||||
|
||||
@@ -8,18 +8,18 @@
|
||||
"language": "zh-CN",
|
||||
"chapters": 62,
|
||||
"images": 20,
|
||||
"bytes": 2670517,
|
||||
"sha256": "d84eca75032f6dca52cfa7a9ce8070e5a8ad13c91b94e7e963d097371d14b16b",
|
||||
"sourceSha256": "d0546cfc6eaa0ac96397f29f3c8b51d3e6e5bcc4c3f84e03d0f7a03854d6eee8"
|
||||
"bytes": 2675205,
|
||||
"sha256": "d1416eccc73fead1ee68044bf34e9ddaeb4a528117f770b0894d80cf90ee7201",
|
||||
"sourceSha256": "c74abe6b9de6181109f43e460d907b9c54ea0cadf11efe28ae8d7b11e083833d"
|
||||
},
|
||||
"en-US": {
|
||||
"file": "life-level-up-guide-en.epub",
|
||||
"language": "en-US",
|
||||
"chapters": 62,
|
||||
"images": 20,
|
||||
"bytes": 2639972,
|
||||
"sha256": "f285bf90d3da67244389f126ccb2a24d08fb63a717cc42cf98cc3394d211b17d",
|
||||
"sourceSha256": "80431ee36a3ef744353c8c7a5b6e7dac1faa3e46b0b5e092468134f726672998"
|
||||
"bytes": 2644546,
|
||||
"sha256": "51323344fe57ee1016a709d4294c889dbb7b9042fd94120ee38cf85e12d1b5da",
|
||||
"sourceSha256": "9e45b09aae8d7fb8314009197b5cdf456dae54cbced796cd8049553581b96e1e"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
Binary file not shown.
File diff suppressed because one or more lines are too long
Binary file not shown.
File diff suppressed because one or more lines are too long
@@ -7,14 +7,14 @@
|
||||
"zh-CN": {
|
||||
"file": "life-level-up-guide-zh.pdf",
|
||||
"language": "zh-CN",
|
||||
"pages": 325,
|
||||
"pages": 330,
|
||||
"chapters": 62,
|
||||
"bytes": 4643471,
|
||||
"sha256": "a91bcb73d8f4d781e98eab41e6c84747113e16bcb6b530173789a0ac2f9ed60d",
|
||||
"sourceEpubSha256": "d84eca75032f6dca52cfa7a9ce8070e5a8ad13c91b94e7e963d097371d14b16b",
|
||||
"semanticSha256": "166a9134169fb1aadcfc6fa109fd2c6addcd1420732d8f542db55b212316405d",
|
||||
"outlineEntries": 753,
|
||||
"linkAnnotations": 766,
|
||||
"bytes": 4666039,
|
||||
"sha256": "c1ea9f2f64d3e377f8993f2c5b696d7d3a8aaec22c472ed33503f3d1a2dd7daa",
|
||||
"sourceEpubSha256": "d1416eccc73fead1ee68044bf34e9ddaeb4a528117f770b0894d80cf90ee7201",
|
||||
"semanticSha256": "bd1bda23d55910dfae5c3269bf2187bc551b7d846930874cb2f3b4b17c3ae6e9",
|
||||
"outlineEntries": 771,
|
||||
"linkAnnotations": 777,
|
||||
"images": 21,
|
||||
"fonts": [
|
||||
{
|
||||
@@ -42,14 +42,14 @@
|
||||
"en-US": {
|
||||
"file": "life-level-up-guide-en.pdf",
|
||||
"language": "en-US",
|
||||
"pages": 367,
|
||||
"pages": 372,
|
||||
"chapters": 62,
|
||||
"bytes": 3845798,
|
||||
"sha256": "38ccb77184b2c47cbd309d70476dcbeda8b2a52ee45d6136fc2edf92d6188f22",
|
||||
"sourceEpubSha256": "f285bf90d3da67244389f126ccb2a24d08fb63a717cc42cf98cc3394d211b17d",
|
||||
"semanticSha256": "162a8bed07dc8384e2d6b6627a3ed09219e97687f2e298d4814259c336e1961b",
|
||||
"outlineEntries": 753,
|
||||
"linkAnnotations": 781,
|
||||
"bytes": 3865628,
|
||||
"sha256": "8448bc28a2070fb734dbd2607c85b1e53a677dcd29711b89335f995401e95787",
|
||||
"sourceEpubSha256": "51323344fe57ee1016a709d4294c889dbb7b9042fd94120ee38cf85e12d1b5da",
|
||||
"semanticSha256": "c578dd7012b4661c765e38ba3f807e826843cedeb89cfda19b1e0c224d56a055",
|
||||
"outlineEntries": 771,
|
||||
"linkAnnotations": 794,
|
||||
"images": 21,
|
||||
"fonts": [
|
||||
{
|
||||
|
||||
@@ -19,6 +19,7 @@ updated: 2026-09-02
|
||||
| **真实问题** | 有具体场景、受影响的人和需要完成的动作，而不是“我想变好” | [认知篇](../threads/part-1/1-understanding.md) |
|
||||
| **场景** | 能力将在何时、何地、对谁被使用 | [CEFR 目标与自测](../threads/part-1/0-cefr.md) |
|
||||
| **完成标准** | 一个别人可以观察、复查或复述的结果 | [AI 任务简报](../templates/ai-task-brief.md) |
|
||||
| **评估集** | 用正常、边界、缺失/冲突和真实脱敏样本检查输出，而不是只展示模型擅长的例子 | [AI 任务简报](../templates/ai-task-brief.md) |
|
||||
| **主问题** | 一周只追踪的一个可回答问题 | [九十日行动篇](../threads/part-5/90-day-plan.md) |
|
||||
|
||||
## 证据与判断
|
||||
@@ -90,7 +91,9 @@ updated: 2026-09-02
|
||||
| **互动修复** | 没听清、范围不明或暂时未知时，请求重复、确认、争取思考时间并总结共识 | [口语篇](../threads/part-1/5-speaking.md) · [求职英语篇](../threads/part-1/8-job-search-english.md) |
|
||||
| **作品** | 能被别人阅读、使用、质疑或改进的输出，不限于产品 | [作品篇](../threads/part-3/4-artifacts-and-delivery.md) |
|
||||
| **交付** | 在真实受众、用户或约束下完成并接受结果 | [九十日行动篇](../threads/part-5/90-day-plan.md) |
|
||||
| **人工门** | AI 可以协助，但事实、权限、隐私、成本和最终判断必须由人确认 | [AI 项目开发](../threads/part-3/2-ai-development-and-resource-layer.md) |
|
||||
| **人工门** | AI 可以协助，但来源、事实、权限、隐私、质量、成本和最终责任必须在发布或执行前由具名的人确认 | [AI 任务简报](../templates/ai-task-brief.md) · [AI 项目开发](../threads/part-3/2-ai-development-and-resource-layer.md) |
|
||||
| **速度债** | 通过缩短即时用时却增加返工、核验、沟通或事故风险而欠下的真实成本 | [AI 学习记录](../templates/ai-learning-log.md) |
|
||||
| **AI 学习记录** | 比较无 AI、辅助、延迟独立和平行任务，记录工具贡献与独立能力的私密工作纸 | [AI 学习记录](../templates/ai-learning-log.md) |
|
||||
| **回滚** | 失败时暂停、降级、切换或恢复到已知可用状态 | [AI 学习与项目实践](../threads/part-3/1-ai-learning.md) |
|
||||
|
||||
## 生活与边界
|
||||
|
||||
Vendored
+123
-45
@@ -1,53 +1,131 @@
|
||||
---
|
||||
title: AI 学习记录模板
|
||||
description: 保存无 AI 基线、引导练习、错误、证据、独立表现和迁移结果，让学习跨会话继续并可以复测。
|
||||
updated: 2026-08-31
|
||||
title: AI 学习记录：从工具协作到独立能力
|
||||
description: 记录 AI 学习切片的真实任务、无 AI 基线、工具范围、来源核验、主动产出、延迟复测、迁移和下一周期决定。
|
||||
updated: 2026-09-02
|
||||
---
|
||||
|
||||
# AI 学习记录模板
|
||||
# AI 学习记录：从工具协作到独立能力
|
||||
|
||||
每次只记录一个可检查的学习切片。AI 生成的摘要不是事实来源，原始作品和来源才是。
|
||||
每次只记录一个可检查的学习切片。AI 生成的摘要、评分和解释不是事实来源；原始材料、自己的首版、实际作品和真实反馈才是证据。不要把学习记录写成“我和模型聊了什么”，而要写成“工具退出后，我还能独立完成什么”。
|
||||
|
||||
## 1. 任务与起点
|
||||
|
||||
```markdown
|
||||
# AI Learning Log — YYYY-MM-DD
|
||||
# AI Learning Log - YYYY-MM-DD
|
||||
|
||||
## 任务
|
||||
- 真实问题：
|
||||
- 完成标准：
|
||||
- 本次最小切片：
|
||||
|
||||
## 基线
|
||||
- 无 AI 尝试：
|
||||
- 暴露的缺口：
|
||||
- 可用材料与来源：
|
||||
|
||||
## AI 协作
|
||||
- AI 做了什么：
|
||||
- 我保留了哪些判断：
|
||||
- 哪些回答需要核验：
|
||||
|
||||
## 主动产出
|
||||
- 关闭 AI 后完成的作品：
|
||||
- 独立完成用时：
|
||||
- 测试、反馈或评分：
|
||||
|
||||
## 三次对照
|
||||
| 样本 | 条件 | 任务完成/质量 | 用时/返工 | 信心 | 证据位置 |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| 无 AI 基线 | | | | | |
|
||||
| AI 辅助版 | | | | | |
|
||||
| 延迟独立版（3–7 天后） | | | | | |
|
||||
|
||||
## 错误与迁移
|
||||
| 错误/缺口 | 证据 | 可能原因 | 平行任务结果 |
|
||||
| --- | --- | --- | --- |
|
||||
| | | | |
|
||||
|
||||
## 下一步
|
||||
- 下一项最小任务：
|
||||
- 完成标准：
|
||||
- 需要的材料：
|
||||
- 应保存的证据：
|
||||
- 交接给谁/何时：
|
||||
- 停止或回滚条件：
|
||||
真实场景与受众：
|
||||
本次主问题：
|
||||
要完成的动作与标准：
|
||||
材料、版本与来源：
|
||||
时间、设备和环境限制：
|
||||
允许的词典 / 翻译 / 搜索 / AI / 人工帮助：
|
||||
隐私、版权、考试与责任边界：
|
||||
```
|
||||
|
||||
## 2. 无 AI 基线
|
||||
|
||||
```markdown
|
||||
独立尝试的开始/结束时间：
|
||||
无 AI 样本位置：
|
||||
我当时能够解释或完成的内容：
|
||||
暴露的错误、未知和判断盲点：
|
||||
当时的信心（0-2）：
|
||||
```
|
||||
|
||||
不要先让 AI 写答案再把它叫作起点。若任务本身允许查词或看文档，记录允许的帮助；基线的价值是说明你在这些条件下独立能做什么。
|
||||
|
||||
## 3. 约定 AI 只做什么
|
||||
|
||||
```markdown
|
||||
模型 / 版本 / 日期：
|
||||
使用模式：诊断 / 追问 / 解释 / 反例 / 转写 / 候选 / 其他
|
||||
AI 可以做：
|
||||
AI 不可以做：
|
||||
传给模型的最小材料：
|
||||
不会传入的字段或文件：
|
||||
```
|
||||
|
||||
每轮只推进一个切片。先让模型指出缺口、假设、来源和不确定性，不先请求整段代做。
|
||||
|
||||
## 4. 交互与来源核验
|
||||
|
||||
| 回合 | 我的问题/动作 | 模型建议 | 建议类型 | 来源/不确定性 | 我的决定 |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| | | | 解释 / 追问 / 候选 / 反馈 / 代码 / 其他 | | 采纳 / 部分 / 拒绝 / 待核验 |
|
||||
|
||||
```markdown
|
||||
我核对过的原始来源：
|
||||
我发现的错误、过时或捏造内容：
|
||||
哪些建议改变了我的理解，为什么：
|
||||
哪些建议只改变了表面表达：
|
||||
```
|
||||
|
||||
来源、事实、权限、隐私和最终判断由人确认。模型说得自信不提高证据等级。
|
||||
|
||||
## 5. 关闭 AI 后主动产出
|
||||
|
||||
```markdown
|
||||
关闭 AI 的时间：
|
||||
独立完成的作品、解释、代码或决定：
|
||||
独立用时：
|
||||
仍然需要的提示：
|
||||
我能否解释每个关键步骤：是 / 部分 / 否
|
||||
测试、读者/用户反馈或运行结果：
|
||||
```
|
||||
|
||||
## 6. 三次对照
|
||||
|
||||
| 样本 | 条件 | 任务/质量 | 用时 | 返工 | 信心 | 证据位置 |
|
||||
| --- | --- | --- | ---: | ---: | ---: | --- |
|
||||
| 无 AI 基线 | | | | | | |
|
||||
| AI 辅助版 | | | | | | |
|
||||
| 延迟独立版（3-7 天） | | | | | | |
|
||||
| 平行任务版（换条件） | | | | | | |
|
||||
|
||||
| 结果 | 更谨慎的解释 |
|
||||
| --- | --- |
|
||||
| 辅助版和独立版都改善 | 可能形成了可撤掉的支架，仍需更多迁移 |
|
||||
| 只有辅助版改善 | 成品改善，关键步骤可能被外包 |
|
||||
| 时间变短但返工增多 | 速度换来了速度债 |
|
||||
| 信心上升但事实/测试下降 | 先校准判断，再增加工具权限 |
|
||||
|
||||
一次对照不能证明因果，但能比“AI 很有用”更清楚地记录工具贡献。
|
||||
|
||||
## 7. 错误、反馈与迁移
|
||||
|
||||
| 错误/缺口 | 样本 | 类型 | 下一次只修什么 | 平行任务结果 |
|
||||
| --- | --- | --- | --- | --- |
|
||||
| | | 事实 / 词汇 / 结构 / 语法 / 策略 / 注意 / 工具依赖 | | |
|
||||
|
||||
```markdown
|
||||
真实读者/用户复述或执行结果：
|
||||
我采纳、部分采纳、拒绝或延后的反馈及理由：
|
||||
换主题/受众/模态/时间后仍能完成的动作：
|
||||
仍然只在原题成功的动作：
|
||||
```
|
||||
|
||||
## 8. 延迟保持与交接
|
||||
|
||||
```markdown
|
||||
第 1 天：撤掉了什么，独立完成什么：
|
||||
第 3-7 天：换了什么条件，结果如何：
|
||||
第 14 天：关闭旧对话后能否解释和使用：
|
||||
第 30 天：真实任务是否减少猜测或返工：
|
||||
状态文件位置：
|
||||
交接给谁、何时、什么权限：
|
||||
停止或回滚条件：
|
||||
```
|
||||
|
||||
## 9. 下一周期决定
|
||||
|
||||
```markdown
|
||||
工具真正帮助的部分：
|
||||
工具制造的风险或依赖：
|
||||
我已经拥有的独立能力证据：
|
||||
仍无证据的结论：
|
||||
继续 / 降档 / 改变变量 / 暂停 / 求助：
|
||||
下一项最小任务：
|
||||
下一次复查日期：
|
||||
```
|
||||
|
||||
相关入口：[认知篇：把努力变成可验证的学习](../threads/part-1/1-understanding.md) ｜ [AI 学习、项目开发与资源层创业](../threads/part-3/1-ai-learning.md) ｜ [AI 任务简报](ai-task-brief.md) ｜ [学习状态](learning-state.md) ｜ [证据链模板](evidence-chain.md)
|
||||
|
||||
Vendored
+133
-39
@@ -1,48 +1,142 @@
|
||||
---
|
||||
title: AI 任务简报模板
|
||||
description: 用真实场景、数据边界、验收标准和人工责任定义一次 AI 学习或项目任务。
|
||||
updated: 2026-08-24
|
||||
title: AI 任务简报：从问题到人工验收
|
||||
description: 在 AI 学习或项目开始前写清真实任务、数据边界、来源、评估集、人工责任、成本、失败处理和交接条件。
|
||||
updated: 2026-09-02
|
||||
---
|
||||
|
||||
# AI 任务简报模板
|
||||
# AI 任务简报：从问题到人工验收
|
||||
|
||||
复制到私有项目目录。不要写入密码、证件、精确住址、医疗隐私、客户资料或未获授权的第三方信息。
|
||||
复制到私有项目目录，在模型调用前填写。它不是 prompt 收集器，而是一张开始许可：如果任务、数据、来源、验收或责任人仍然不清楚，就先不要让模型生成最终结果。
|
||||
|
||||
不要写入密码、证件、精确住址、医疗隐私、儿童数据、客户记录、未公开漏洞或无权使用的第三方内容。敏感材料先脱敏，或改用组织批准的环境。
|
||||
|
||||
## 1. 任务与责任
|
||||
|
||||
```markdown
|
||||
# AI Task Brief
|
||||
# AI Task Brief - YYYY-MM-DD
|
||||
|
||||
更新时间：YYYY-MM-DD
|
||||
|
||||
## 任务
|
||||
- 真实场景：
|
||||
- 使用者/受众：
|
||||
- 要完成的决策或动作：
|
||||
- 截止时间：
|
||||
|
||||
## 输入与边界
|
||||
- 已知事实与来源：
|
||||
- 允许使用的文件：
|
||||
- 数据敏感等级：公开 / 内部 / 机密 / 受限
|
||||
- 明确不提供的材料：
|
||||
- 版权或授权限制：
|
||||
|
||||
## 输出与验收
|
||||
- 最终交付物：
|
||||
- 格式与长度：
|
||||
- 可观察的完成标准：
|
||||
- 必须由人确认的事项：
|
||||
|
||||
## 工作方式
|
||||
- AI 可以做：
|
||||
- AI 不可以做：
|
||||
- 人工审阅者：
|
||||
- 失败时如何暂停、回滚或通知：
|
||||
|
||||
## 证据位置
|
||||
- 无 AI 基线：
|
||||
- 原始来源：
|
||||
- 测试/反馈：
|
||||
- 最终版本：
|
||||
真实场景：
|
||||
使用者/受众：
|
||||
要完成的决策或动作：
|
||||
为什么现在做：
|
||||
截止时间与不可错过的节点：
|
||||
最终负责人与人工审阅者：
|
||||
如果不做或延迟，谁会受影响：
|
||||
```
|
||||
|
||||
验收标准要写成动作，例如“用户能在 10 分钟内完成导入并看到可解释的错误报告”，不要写“体验好”或“足够智能”。
|
||||
把“学 AI”“做一个智能助手”改成可观察的动作，例如“用户能在 10 分钟内导入一份文件，并看到一份能解释来源和错误的报告”。
|
||||
|
||||
## 2. 输入、来源与数据边界
|
||||
|
||||
| 输入/主张 | 类型 | 来源、版本与位置 | 可否提供给模型 | 核验人 | 失效条件 |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| | 事实 / 推断 / 个人经验 / 用户数据 / 第三方内容 | | 公开 / 脱敏 / 组织批准 / 禁止 | | |
|
||||
|
||||
```markdown
|
||||
允许使用的文件和字段：
|
||||
明确不提供的材料：
|
||||
数据敏感等级：公开 / 内部 / 机密 / 受限
|
||||
收集、传输、留存和删除期限：
|
||||
是否取得必要同意：
|
||||
版权、许可证、引用和署名要求：
|
||||
模型能接触的最小数据范围：
|
||||
```
|
||||
|
||||
来源不等于真相。每个关键主张要能回到原文、版本、数据口径或直接观察；模型给出的链接、数字和引语仍需人工核验。
|
||||
|
||||
## 3. 输出与评估集
|
||||
|
||||
```markdown
|
||||
最终交付物：
|
||||
格式、长度和受众：
|
||||
可观察的完成标准：
|
||||
必须通过的硬门：
|
||||
允许的可接受变体：
|
||||
明确不能出现的内容：
|
||||
```
|
||||
|
||||
建立一份小而真实的评估集，不要只用模型最擅长的示例：
|
||||
|
||||
| 样本 | 输入条件 | 预期结果 | 不可接受结果 | 实测结果 | 证据位置 |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| 正常样本 | | | | | |
|
||||
| 边界样本 | | | | | |
|
||||
| 缺失/冲突输入 | | | | | |
|
||||
| 真实历史样本（已脱敏） | | | | | |
|
||||
|
||||
验收至少回答：事实是否可追溯，关键动作是否完成，错误是否可见，权限是否正确，失败是否可停止。流畅、快速或“像人”不是验收标准。
|
||||
|
||||
## 4. AI 的工作范围
|
||||
|
||||
```markdown
|
||||
AI 可以做：提问 / 分类 / 候选解释 / 反例 / 转写 / 草稿 / 测试建议
|
||||
AI 不可以做：最终事实判断 / 虚构来源 / 代替责任人批准 / 越过权限
|
||||
/ 完成禁止外部帮助的考试或申请 / 自动发布
|
||||
使用模式：诊断 / 辅助 / 生成候选 / 批量处理 / 其他
|
||||
模型、版本、区域和调用日期：
|
||||
提示词、系统指令或工作流版本位置：
|
||||
```
|
||||
|
||||
先让模型复述目标、输入、限制、未知项和验收，再开始生成。提示词版本化；不要把一次聊天窗口当作唯一的项目记录。
|
||||
|
||||
## 5. 人工门
|
||||
|
||||
| 门 | 必须由谁确认 | 通过证据 | 未通过时 |
|
||||
| --- | --- | --- | --- |
|
||||
| 来源门 | | 原始链接、版本、位置 | 标记未核验，不传播 |
|
||||
| 事实门 | | 抽样核对、数据口径 | 删除或降级主张 |
|
||||
| 隐私/权限门 | | 数据范围、同意、访问日志 | 停止处理并脱敏 |
|
||||
| 质量门 | | 评估集、边界样本、真实反馈 | 修复、缩小或拒绝交付 |
|
||||
| 成本门 | | token、时间、人工返工与预算 | 降级模型或停止 |
|
||||
| 责任门 | | 具名批准和披露文字 | 不发布、不自动执行 |
|
||||
|
||||
关键决定不能只由生成模型、自动评分或单个开发者自己批准。高风险领域回到当前一手材料和合格专业人员。
|
||||
|
||||
## 6. 成本、留存与可逆性
|
||||
|
||||
```markdown
|
||||
预计调用量、时间和费用上限：
|
||||
人工审阅与返工预算：
|
||||
数据是否进入训练、日志或第三方留存：
|
||||
可接受延迟与降级方案：
|
||||
可以撤回、重跑或恢复到哪个版本：
|
||||
发布范围与灰度对象：
|
||||
```
|
||||
|
||||
一次便宜的调用如果造成隐私事件、错误决定或大量返工，真实成本并不便宜。记录模型费用之外的人工时间、审阅、失败、重跑和沟通成本。
|
||||
|
||||
## 7. 失败、暂停与回滚
|
||||
|
||||
| 触发条件 | 立即动作 | 通知谁 | 恢复/回滚位置 |
|
||||
| --- | --- | --- | --- |
|
||||
| 找不到来源或版本 | 停止传播，回到原始材料 | | |
|
||||
| 评估集硬门失败 | 禁止发布或自动执行 | | |
|
||||
| 输入越过权限边界 | 停止上传，撤销访问 | | |
|
||||
| 成本/延迟超过上限 | 降级、限流或停止 | | |
|
||||
| 真实用户报告伤害 | 下线、保留证据、升级处理 | | |
|
||||
|
||||
如果没有明确的停止人、通知路径和已知可用版本，就还没有准备好接入真实流程。
|
||||
|
||||
## 8. 交接与公开说明
|
||||
|
||||
```markdown
|
||||
当前状态：未开始 / 实验 / 内部试用 / 灰度 / 已交付 / 已停止
|
||||
已完成与证据位置：
|
||||
未决问题和风险：
|
||||
下一项最小任务：
|
||||
交接负责人、时间和权限：
|
||||
读者/用户需要知道的 AI 参与范围：
|
||||
来源、隐私、版权和利益关系说明：
|
||||
```
|
||||
|
||||
交接者不应需要翻找聊天记录才能知道下一步。公开文本至少说明工具参与了哪一步、哪些事实由人确认、哪些内容仍是候选。
|
||||
|
||||
## 9. 开始前检查
|
||||
|
||||
- [ ] 真实受众、动作、截止时间和负责人已写清；
|
||||
- [ ] 输入字段、敏感等级、同意、版权和留存边界已确认；
|
||||
- [ ] 关键来源、版本和失效条件可回溯；
|
||||
- [ ] 正常、边界、冲突和历史脱敏样本已进入评估集；
|
||||
- [ ] AI 可以做与不可以做的范围已写清；
|
||||
- [ ] 人工门、费用上限、停止条件和回滚位置已指定；
|
||||
- [ ] 交接、披露和下一次复查日期已确定。
|
||||
|
||||
@@ -84,7 +84,7 @@ sources_checked: 2026-08-24
|
||||
|
||||
## 一页任务简报
|
||||
|
||||
复制到自己的私有项目目录，敏感资料先脱敏，不把密码、证件、客户记录或第三方隐私直接交给通用模型。
|
||||
先复制完整的[AI 任务简报](../../templates/ai-task-brief.md)到自己的私有项目目录。下面是最小字段；敏感资料先脱敏，不把密码、证件、客户记录或第三方隐私直接交给通用模型。
|
||||
|
||||
```markdown
|
||||
# AI Task Brief
|
||||
@@ -155,7 +155,7 @@ AI 不可以做：
|
||||
|
||||
## 三次对照：证明 AI 帮了什么
|
||||
|
||||
同一个任务至少留下三份样本：
|
||||
用[AI 学习记录](../../templates/ai-learning-log.md)为同一个任务至少留下三份样本：
|
||||
|
||||
1. **无 AI 基线**：先独立完成，记录用时、质量、卡点和信心；
|
||||
2. **AI 辅助版**：只让 AI 做约定的工作，记录提示、来源、采纳与拒绝的建议，以及返工；
|
||||
|
||||
@@ -419,6 +419,33 @@ test("learning principles turn effort into a complete evidence loop", () => {
|
||||
}
|
||||
});
|
||||
|
||||
test("AI work papers keep evaluation, human gates, and independent transfer visible", () => {
|
||||
const cases = [
|
||||
{
|
||||
brief: "templates/ai-task-brief.md",
|
||||
log: "templates/ai-learning-log.md",
|
||||
briefTerms: ["输入、来源与数据边界", "输出与评估集", "人工门", "失败、暂停与回滚", "开始前检查"],
|
||||
logTerms: ["无 AI 基线", "交互与来源核验", "三次对照", "延迟保持与交接", "下一周期决定"],
|
||||
},
|
||||
{
|
||||
brief: "en/templates/ai-task-brief.md",
|
||||
log: "en/templates/ai-learning-log.md",
|
||||
briefTerms: ["Inputs, Sources, and Data Boundary", "Output and Evaluation Set", "Human Gates", "Failure, Pause, and Rollback", "Preflight Check"],
|
||||
logTerms: ["Unaided Baseline", "Interaction and Source Verification", "Three Comparisons", "Delayed Retention and Handover", "Next Cycle Decision"],
|
||||
},
|
||||
];
|
||||
|
||||
for (const { brief, log, briefTerms, logTerms } of cases) {
|
||||
const briefText = readFileSync(resolve(process.cwd(), "docs", brief), "utf8");
|
||||
const logText = readFileSync(resolve(process.cwd(), "docs", log), "utf8");
|
||||
for (const term of briefTerms) expect(briefText, brief).toContain(term);
|
||||
for (const term of logTerms) expect(logText, log).toContain(term);
|
||||
expect(briefText).toContain("AI Task Brief");
|
||||
expect(logText).toContain("AI Learning Log");
|
||||
expect(logText).toContain("evidence-chain.md");
|
||||
}
|
||||
});
|
||||
|
||||
test("writing turns tool polish into accountable revision and delivery", () => {
|
||||
const cases = [
|
||||
{
|
||||
@@ -879,6 +906,10 @@ test("heading-only search keeps long-form chapters and tools discoverable withou
|
||||
await expect(zhSearchBox.getByRole("link", { name: /认知篇：把努力变成可验证的学习/ }).first()).toBeVisible();
|
||||
await zhSearchBox.locator("input").fill("英语能力诊断：四项基线与迁移记录");
|
||||
await expect(zhSearchBox.getByRole("link", { name: /英语能力诊断：四项基线与迁移记录/ }).first()).toBeVisible();
|
||||
await zhSearchBox.locator("input").fill("AI 任务简报：从问题到人工验收");
|
||||
await expect(zhSearchBox.getByRole("link", { name: /AI 任务简报：从问题到人工验收/ }).first()).toBeVisible();
|
||||
await zhSearchBox.locator("input").fill("AI 学习记录：从工具协作到独立能力");
|
||||
await expect(zhSearchBox.getByRole("link", { name: /AI 学习记录：从工具协作到独立能力/ }).first()).toBeVisible();
|
||||
await zhSearchBox.locator("input").fill("写作篇：从初稿到可验证修订");
|
||||
await expect(zhSearchBox.getByRole("link", { name: /写作篇：从初稿到可验证修订/ }).first()).toBeVisible();
|
||||
await zhSearchBox.locator("input").fill("写作证据卡：从工具润色到署名交付");
|
||||
@@ -922,6 +953,10 @@ test("heading-only search keeps long-form chapters and tools discoverable withou
|
||||
await expect(enSearchBox.getByRole("link", { name: /Learning Principles: Turn Effort into Verifiable Learning/ }).first()).toBeVisible();
|
||||
await enSearchBox.locator("input").fill("English Diagnostic: Four-skill Baseline and Transfer Record");
|
||||
await expect(enSearchBox.getByRole("link", { name: /English Diagnostic: Four-skill Baseline and Transfer Record/ }).first()).toBeVisible();
|
||||
await enSearchBox.locator("input").fill("AI Task Brief: From Problem to Human Acceptance");
|
||||
await expect(enSearchBox.getByRole("link", { name: /AI Task Brief: From Problem to Human Acceptance/ }).first()).toBeVisible();
|
||||
await enSearchBox.locator("input").fill("AI Learning Log: From Tool Collaboration to Independent Ability");
|
||||
await expect(enSearchBox.getByRole("link", { name: /AI Learning Log: From Tool Collaboration to Independent Ability/ }).first()).toBeVisible();
|
||||
await enSearchBox.locator("input").fill("Writing: From Draft to Verifiable Revision");
|
||||
await expect(enSearchBox.getByRole("link", { name: /Writing: From Draft to Verifiable Revision/ }).first()).toBeVisible();
|
||||
await enSearchBox.locator("input").fill("Writing Evidence Card: From Tool Polish to Accountable Delivery");
|
||||
|
||||
Reference in New Issue
Block a user