[defect·copy·P1] The memory-continuity line names neither what the tracked number counts nor which direction is good

[atomchung]

Status

Blocked — implementation complete; owner-live weekly-return acceptance pending.

Milestone

M1 acceptance hardening — returning-user memory delivery.

Owning outcome

A returning user can understand what the remembered weekly metric counts, whether it improved/worsened/stayed flat against the stored engine goal, and that the milestone count means completed reviews.

Current evidence

• PR fix(card): clarify weekly continuity metrics #730 merged in main@9751ced7671c5a5bbb5a1fefd32bd41dfb6d4f2a.
• Reconciliation labels and direction remain engine-owned. Legacy or unknown metric semantics render only the prior commitment, without a crash or invented meaning/direction.
• en, zh-TW, and zh-CN are corpus-rendered and pinned; the corpus covers 720 rendered surfaces and 190 copy keys.
• The focused tests, Python 3.12 offline suite, mutation probe, independent review, and Python 3.11/3.12 merge-tree CI passed.

These are deterministic implementation facts, not owner-live acceptance.

Remaining scope

On the frozen candidate above, run the owner-observed returning-user weekly review. The owner must be able to explain the remembered metric and whether its movement is good or bad without repository knowledge. Record the result through #713, #610, #676, and #603.

Non-goals

• No new memory architecture, rationale/thesis/profile stream, or card redesign.
• No [feat·M1] First-review value preview — show the engine finding before required questions #714, README, feat(rationale): a canonical stream for why the user says they still hold a position (refs #403, refs #450) #661, [feat·M2] Append-only rationale refresh anchored to a position cycle #403, or weekly-market-read work.
• No metric arithmetic, rule ranking, canonical-state, or persistence change.

Dependencies

• [acceptance·P0] M1 vertical-slice reliability gate — repeatable completion before more capability #713's frozen-SHA reliability sequence.
• [QA·dogfood] Judgment-quality dogfood — does a real consider answer now carry a supported insight? #610, [acceptance·P1] Global FOMO Kernel voice contract — merged, awaiting owner-live verdict #676, and [design·perf] Response-time budget by route — separate real-user latency from QA/test overhead #603 owner-observed evidence.

Stop condition

Close only after the owner-observed weekly-return acceptance passes on the frozen candidate (or a later re-frozen candidate after a recorded failure).

[atomchung]

Owner disposition — bounded implementation is suitable tonight

This is one of today's unresolved owner-visible QA findings and remains a small presentation correction over existing engine-owned memory state.

Implement as its own narrow PR after the active #727 research-aware branch is reviewable:

• derive the localized tracked-quantity label from existing metric/dimension semantics;
• state improved / worsened / unchanged against the existing engine-owned goal direction;
• replace ambiguous “completed records” copy with “completed reviews” in en, zh-TW, and zh-CN;
• keep card sections/length, rule identity, calculations, persistence, recovery and legacy-session compatibility unchanged;
• add mutation coverage for removed/flipped direction and locale alignment;
• run focused card/reconciliation tests and the full suite.

No owner-live walkthrough is required tonight; product acceptance remains for tomorrow. Merge requires independent review and green CI, and any merged product change means tomorrow's acceptance candidate must be re-frozen.

[atomchung]

Routing update from the 2026-08-02 maintenance scan:

• This remains the second and final bounded repair in the pre-owner-live wave.
• Start only after [defect·P1] A run the owner judged a failure cannot be archived — so the dogfood ledger's pass rate is structurally 100% #681 reaches an independently reviewable, CI-green boundary; do not run both implementations in parallel.
• Base the work on the then-current main (currently 8a50b01979de4e1899ae6d5712258a4279025172), not the earlier 5ed29d9 reliability candidate.
• Preserve the existing acceptance boundary: presentation/copy only across en, zh-TW, and zh-CN; no arithmetic, rule identity, canonical state, section, schema, or persistence change.
• After this leaf lands, freeze the exact final SHA and refresh [acceptance·P0] M1 vertical-slice reliability gate — repeatable completion before more capability #713 deterministic evidence before owner-live.

#714 remains deferred until after owner-live; do not couple first-review opening_value or answer_effect into this PR.

[atomchung]

Status update: PR #730 is the sole #682 implementation candidate, based on main@45392276a90ff72f2a34fecca0351882ca779420. It carries only localized continuity rendering/copy and the regenerated deterministic corpus; local full offline suite passed. Python 3.11/3.12 CI is pending. This is not owner-live acceptance and #714 remains deferred.