Grand dashboard: memory-lifecycle-v2, retrieval-quality-v2 and token-efficiency-wiring rows after the GPT-6 repair round

[seathatflowsinourveins]

Summary

This PR updates three foundation gate rows in the shared hot file observability/grand-dashboard/state.json, following the GPT-6 repair round. Each row states only what its merged evidence shows. The two memory and retrieval rows were moved out of their repair PRs because this file is shared; the token-efficiency row replaces the 2026-09-25 checkpoint text. recorded_at_utc moves to 2026-09-26T12:19:40Z, after all cited evidence.

Gate	Before	After	Evidence (merged)
memory-lifecycle-v2	89 of 89 checks on 2.3.2 and 2.4.0 (a withdrawn claim; see the #292 correction)	local integration 82+19 checks pass on 3 builds / upstream all-targets suites pass at 3 tags / swept pages unreadable / gate stays open	blueprints/memory-lifecycle-v2/results-20260926.json (#330)
retrieval-quality-v2	nDCG@10 0.714 vs 0.223, CI 0.367–0.618 (defective runner)	repaired rerun nDCG@10 0.728 vs 0.223, CI 0.381–0.633, decision unchanged	blueprints/retrieval-quality-v2/results-20260926T024558Z.json (#335)
token-efficiency-wiring	4 host tools ahead of pins / 7 without Linux pins	12 of 14 versions match pins / 2 unchecked (npm metadata) / probes fail closed	evidence/artifacts/token-coverage-20260926/ (#333); the counts come from coverage-20260926T064252Z.json

Lane

This is lane:shared because it changes what a shared hot file says (docs/lanes.md). It needs the trading lane's acknowledgement, as a comment or review, before merge. Only these three foundation rows and the timestamp change; no trading row is touched.

Verification

• tests.test_grand_dashboard, tests.test_memory_lifecycle_v2 and tests.test_retrieval_quality_v2: 99 OK. DashboardCheckpointTests now checks that the checkpoint is not older than the memory results it cites.
• scripts/validate.py, evidence_manifest.py --check and the other validators: all pass.
• Every state and title matches the renderer's token alphabet [A-Za-z0-9_. /+()-]{1,160}.

SOTA sources

• Repository lane and hot-file protocol: docs/lanes.md
• Evidence classes behind each row: docs/acceptance-evidence-policy.md

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[seathatflowsinourveins]

Trading-lane acknowledgement requested (lane:shared, per docs/lanes.md). This PR changes only three foundation rows (memory-lifecycle-v2, retrieval-quality-v2, token-efficiency-wiring) and recorded_at_utc in observability/grand-dashboard/state.json. No trading row changes. Each row cites evidence already merged on main (#330, #335, #333).

[seathatflowsinourveins]

Trading-lane ack (lane:shared), 2026-09-26.

Diffed against the merge base with main. The PR modifies only three foundation gates (memory-lifecycle-v2, retrieval-quality-v2, token-efficiency-wiring) plus recorded_at_utc. No lane, worker or trading gate changes.

Checked at the PR head:

• all three states and titles match the 160-character token pattern;
• each evidence_ref exists;
• tests.test_grand_dashboard passes (15 OK).

The numbers are the foundation lane's, and I didn't re-derive them. Merge on green CI.

🤖 Generated with Claude Code