P328 · Evaluation & feedback

Qualify observations by when they were captured

Distinguish historical evidence, unverified snapshots, current observations and unknown state.

Editorially reviewed

These examples and illustrative results are independently authored teaching materials, not measured model results.

Use case

Investigate which skill influenced last month’s session. Teaching materials include a captured body/digest and today’s HEAD. Current checkout cannot authenticate a past installation, and missing original paths do not erase preserved copies.

Mechanism

Scope claims by time and label historical evidence, unverified snapshot, current observation or unknown. Use captures/archives for past claims and present checkout for now; record retrieved_at separately from as_of. Verify copy provenance/session binding and leave uncaptured installation versions unknown. Cached data obtained today remains uncertain if observation time is unknown.

Bad example

Use today’s HEAD as last month’s version and call a retrieved old cache current verified state.

Good example

Use captured body/hash for that historical file, label today’s HEAD separately and leave whole-install version unknown. Keep retrieval and observation times distinct and verify archive provenance without generalizing one file to the environment.

Why the change matters

Time forms part of evidence meaning. Separating acquisition from observation prevents old information becoming current by relabeling, and file evidence authenticates only its object.

Observable expectation

Teaching records name session date, captured body, present checkout and unknown installation fields. Empty as_of with today’s retrieved_at cannot automatically become fresh. Inspect archives despite lost paths; unknown origin remains unverified.

Limits

Timestamps/digests do not establish truth or environment-wide identity. Cache, local patches and timezones can differ. Limit historical conclusions rather than inventing the past from present state.

Sources and evidence

Read the editorial criteria