AI-ASSISTED DISCOVERY: READING RECORD
Companion guide: https://yenra.com/sci/ai.html
Created October 15, 2026. Copy this blank record for each research claim.
Purpose: distinguish what AI proposed from what researchers established.
Use public research information; leave private data and credentials out.

Paper title, authors and DOI/URL:
Publication date and version checked:
Corrections, updates or retractions checked on:

1. QUESTION
What scientific question did the team ask?
What would count as progress?

2. CONTRIBUTIONS
What did the AI produce (prediction, design, hypothesis, code, ranking)?
What did people supply, decide, measure and interpret?
Which model, data and tools were used?

3. EVIDENCE
What was actually tested, and against what comparison?
Was evaluation separate from training and candidate selection?
How many candidates were proposed, tested, successful and repeated?
What uncertainty, failure cases and selection criteria were reported?
Is the evidence computational, experimental, independently replicated,
or demonstrated in an intended application?
For mathematics: what construction or proof was checked, and how?

4. INTERPRETATION
Write the narrowest conclusion supported by the evidence:
What remains unresolved?
What observation would weaken or overturn the conclusion?
What follow-up would make the result more useful?

5. RECHECK
Code, data, methods and supporting information:
Independent follow-up or replication:
Next reason to revisit this record:

An AI confidence score, a persuasive explanation, and agreement between
AI agents are inputs to evaluation. Record the external check that tests
the claim. A result in a laboratory or benchmark has its stated scope.