Skip to content

Reading Evidence Card: From Finished Summary to Real Delivery

Finishing a text is an action, not proof of comprehension. Copy this card into a private note for one article, chapter, documentation section, or source set at a time. Do not record passwords, client data, unauthorised third-party information, or unnecessary personal details.

1. Define the Task, Reader, and Boundary

markdown
# Reading Evidence - YYYY-MM-DD

Real situation:
Task: locate information / follow an argument / read documentation / build stamina / synthesise sources / support a decision
Who must know, decide, execute, or continue something after the reading?
Completion standard:
Time limit:
Dictionary / translation / search / notes / AI allowed:
Copyright, privacy, and retention boundary:

Write completion as a result, such as "can run the minimum example under the current version", not only "finished".

2. Record Source and Version

ItemMain sourceParallel sourceComparable?
Author/maintainer, title, link
Edition, publication date, or version
Access date and length
Author/institution position or interest
Quotation, image, code, and copyright boundary
Topic, genre, and difficulty

Web pages change, become paid, or disappear. Keep necessary citation details; do not copy an entire article without permission.

3. Preserve an Unpolished First Pass

markdown
First-pass conditions:
First-pass time:
Did I look up every word?
One-sentence gist:
How the text moves:
Three certain details:
What the author/maintainer wants me to believe or do:
One unresolved point:
Action I would take from this understanding:

Finish the first pass before checking words or explanations. Do not replace it with a polished translation.

4. Build a Six-Layer Barrier Map

Paragraph/pageWhere I got stuckBarrier layerTask impactThis cycle's action
vocabulary/grammar / structure/reference / background/schema / argument/evidence / layout/technical / attention/capacitymeaning-changing / task-blocking / recurring / effortful / low-impact
markdown
One to three barriers that most affect the task:
Low-impact words or sentences I will skip for now:
Possible device, version, fatigue, or environment effects:

If almost every sentence blocks the gist, choose easier parallel material or learn domain chunks first. Lowering the entry point does not lower the destination.

5. Map Claims, Evidence, and Limits

Paragraph/pageFunctionAuthor claimSource evidence/exampleLimit, counterexample, or unknown
markdown
Facts stated directly:
Inference supported by the source:
My own judgment or association:
Evidence still missing:
What new evidence would change my judgment:

Separate what the author says from whether you agree. For technical, health, legal, or financial claims, return to reliable primary sources or qualified professional advice.

6. Technical Documentation Fact Card

For an API, SDK, migration guide, command-line help, error reference, or README, add:

markdown
Goal and applicable version:
Prerequisites, permissions, and environment:
Minimum input:
Expected output:
Actual output:
Defaults, ranges, order, or side effects:
Failure condition/error:
Limit explicitly stated in the source:
Part I am still inferring:
Smallest next test:

Reading documentation without running, reasoning through, reproducing, or confirming one minimum fact shows that you saw the page. It does not show mastery.

7. Record Unknown Words, Word Families, and Translation

Word/expressionRole in the taskDefinition/family/collocation checkedSource locationExplainable after tools close?
core / recurring / technical / low-impact
markdown
Background the translation helped me supply:
English structure, stance, or qualifier the translation may hide:
Word I decided not to check yet:

Unknown-word count diagnoses source fit; it does not directly set a level. Word families help guessing and coverage, but a derived form is not automatically usable.

8. Compare Sources and Deliver an Output

markdown
Shared question:
Definition, date, and evidence in source A:
Definition, date, and evidence in source B:
Omitted evidence or institutional interest:
Which differences are factual, and which are values?
OutputAudienceWhat I deliveredAudience responseTraceable basis?
Confirmation / minimum test / explanation / decision memo / handover

After reading, let a reader, code, decision, or next task test the understanding.

9. Record Tools and Feedback

markdown
Tool / reviewer / version and date:
Did it see the source, my first pass, or my summary?
Source locations it says I missed:
Did it confuse fact, inference, and style?
Teacher/peer/primary-source check:
What I accepted/rejected, and why:

AI may extract candidate terms, ask structural questions, and create parallel tasks, but it may invent summaries, citations, background, and author intention. Preserve the first pass before asking it to locate evidence.

10. Test Delayed Retention and Transfer

DateWhat is removedChanged conditionOutputResult
Day 1No full translation or AI summaryMain sourceFirst pass and five-sentence reconstruction
Days 3-7Old notes and dictionary removedParallel sourceGist, evidence, and inference
Day 14Recent feedback removedNew topic or genreExplanation, test, or decision memo
Day 30All prompts closedReal document, report, or handoverComplete delivery
markdown
Method still available after seven days:
Barrier that returned after the source changed:
Did word-by-word translation or summary dependence decrease?
Which post-reading action remains unstable?
One variable for the next cycle:

11. Score and Decide

Score 0-2: 0=not completed, 1=prompt-dependent or unstable, 2=completed under the current condition.

ItemDay 1Day 7Day 14Evidence reason
Preserve the first-pass main line
Separate claim, evidence, and inference
Locate the barrier layer and choose an action
Verify a technical fact or source location
Reconstruct after tools close
Deliver after source, topic, or genre changes
markdown
Current evidence supports: continue main source / downgrade to extensive reading / replace source / move to another problem
Most important evidence:
Conclusion I still cannot draw:
Smallest next task:
Next review date:

Related chapters: Reading: From Word-by-word Translation to Claims and Evidence | Vocabulary | Grammar | Listening | Learning English with AI | Evidence Chain Template

Content CC BY-NC 4.0; site and tooling code MIT.