Decade-first triage mode #116

New issue

Open

opened 2026-07-28 06:06:06 +00:00 by claude-bot · 0 comments

claude-bot commented

2026-07-28 06:06:06 +00:00

Severity: MEDIUM

The idea

A rapid first-pass mode: the photo fills the screen, and one keystroke per photo assigns a decade —
5/6/7/8/9 for the decade, u for unknown, j for junk. Roughly 2-3 seconds per photo, producing
decade-precision evidence or needs_review decisions.

Why it fits

The range-plus-precision data model was built for exactly this. Coarse human bounds everywhere
beat precise dates somewhere: they cap the AI prior, seed constraint propagation, and make the
histogram meaningful early. It also gives a psychologically completable first lap over a collection
that would otherwise feel endless.

Scope

A Triage queue mode in the workspace writing decade-precision evidence marked
source=manual, source_subtype=triage.
Photos still get full review later; triage evidence is a bound, not a decision.
j routes to the photo-exclusion path (#106).
Preload the next image aggressively — at this speed, any latency is the bottleneck.

Honest assessment

Overlaps with what keyboard shortcuts (#45) plus good queue navigation already provide; the genuine
increment is the one-keystroke coarse-decision semantics. Worth it if you expect to personally do a
full pass; skippable if AI plus propagation prove accurate enough that most photos arrive with a
good suggestion.

Done when

A reviewer can triage at roughly one photo per 2-3 seconds
Triage output is distinguishable from considered decisions
Triaged photos still enter normal review
Image preloading keeps pace

References

docs/circa-spec.md (range and precision model)

Depends on: #95, #45, #106.

## Severity: MEDIUM ## The idea A rapid first-pass mode: the photo fills the screen, and one keystroke per photo assigns a decade — 5/6/7/8/9 for the decade, `u` for unknown, `j` for junk. Roughly 2-3 seconds per photo, producing decade-precision evidence or `needs_review` decisions. ## Why it fits The range-plus-precision data model was built for exactly this. Coarse human bounds *everywhere* beat precise dates *somewhere*: they cap the AI prior, seed constraint propagation, and make the histogram meaningful early. It also gives a psychologically completable first lap over a collection that would otherwise feel endless. ## Scope - A Triage queue mode in the workspace writing decade-precision evidence marked `source=manual`, `source_subtype=triage`. - Photos still get full review later; triage evidence is a bound, not a decision. - `j` routes to the photo-exclusion path (#106). - Preload the next image aggressively — at this speed, any latency is the bottleneck. ## Honest assessment Overlaps with what keyboard shortcuts (#45) plus good queue navigation already provide; the genuine increment is the one-keystroke coarse-decision semantics. Worth it if you expect to personally do a full pass; skippable if AI plus propagation prove accurate enough that most photos arrive with a good suggestion. ## Done when - [ ] A reviewer can triage at roughly one photo per 2-3 seconds - [ ] Triage output is distinguishable from considered decisions - [ ] Triaged photos still enter normal review - [ ] Image preloading keeps pace ## References - `docs/circa-spec.md` (range and precision model) Depends on: #95, #45, #106.

claude-bot added this to the v0.4.0 milestone

2026-07-28 06:06:06 +00:00

claude-bot added the

enhancement

frontend

labels

2026-07-28 06:06:06 +00:00