Reading Status
Reading status keeps paper notes honest. It separates raw candidates from verified notes and prevents rough agent output from looking like a finished review.
Status Values
candidate: found by search or agent brief; not yet reviewed.triaged: relevant enough to keep; metadata still may need checking.reading: currently being read; claims are incomplete.verified: metadata, method, evaluation, and limits have been checked.synthesis: mature enough to support a post or research map.archived: not central, superseded, or kept only as context.
Evidence Required
| Status | Required evidence |
|---|---|
candidate | source link or identifier |
triaged | route, topic relevance, and reason to keep |
reading | metadata checked and open questions listed |
verified | claims tied to evidence table, benchmark card, and limitations |
synthesis | concept updates or post/project links exist |
archived | reason for archive is stated |
Status is not a progress label for effort spent. It is a claim about evidence quality.
Promotion Rules
- Candidate to triaged: topic relevance is clear and Paper triage has a destination.
- Triaged to reading: source link and basic metadata are present.
- Reading to verified: claims are tied to evidence, metrics, and limitations.
- Verified to synthesis: the paper changes how a concept, method, or workflow is explained.
Demotion Rules
Demote a note when later inspection weakens the evidence:
| From | To | Trigger |
|---|---|---|
| verified | reading | key metric or split was misread |
| synthesis | verified | synthesis claim is broader than evidence |
| triaged | archived | off-scope or superseded |
| reading | candidate | metadata/source cannot be verified |
Demotion is not failure; it keeps public notes honest.
Checks
- Does the note show what is verified and what is still uncertain?
- Are agent-generated summaries reviewed before promotion?
- Is the status consistent with the amount of evidence in the note?
- Are unpublished results or private project context excluded?
- Would a reader know which claims are safe to reuse?
- Is
to verifyused instead of invented metadata or metrics?