V6 Remediation Log — 2026-06-12

Source of findings: V1_V7_PLAN_SET_AUDIT_2026-06-12.md (V6 sections, §3.3, §5 "V6", §8). Scope: V6 planning corpus only (V6/ docs and gate artifacts). Honesty model: V3/VERIFICATION_AUDIT_2026-05-31.md ("structural gates ≠ operational proof") and the V5 completion-state banner (V5/V5_TODOS.md:40-54).

Checkbox legend used below: [x]→[~] = not locally actionable (external vendor approvals / real-world operations / destination-realm liveness); [x]→[ ] = claimed but contradicted by repo state. No checkbox was marked [x] by this remediation. Every re-mark was judged individually; no batch edits.

1. Broken anchor in gate artifact (audit: V6 finding "non-existent anchor")#

• V6/performance/latency-budgets.v6perf.json:6 — V6/V6_features.md#Performance-Budget-and-Scale → V6/V6_features.md#performance-budgets. The real heading is ## Performance Budgets (V6/V6_features.md), slug performance-budgets.

2. Cost-figure canonicity contradiction#

• V6/V6_features.md §"Concrete Budget Shape": now states V6_DEPENDENCIES.md §19 is the sole canonical cost-figure source, notes arch holds no figures and defers to deps§19, and marks features' own duplicated budget bullets as a non-normative convenience mirror of deps§19 (deps wins on disagreement).
• V6/V6_ARCHITECTURE.md:1150-1151 (defers to DEPENDENCIES) and V6/V6_DEPENDENCIES.md:402-403 (claims canonicity) already agreed; only features was edited for this finding.

3. Lachesis cap mismatch (docs 150–400 "no hard cap" vs artifact 400/400)#

One coherent story adopted: design range 150–400 per resident world; hard cap 400 (the upper design point is the hard cap).

• V6/V6_features.md concurrency-caps bullet: "design target of 150–400 agents" → "design range of 150–400 agents per resident world with a hard cap of 400".
• V6/V6_DEPENDENCIES.md §19 concurrency-caps table, Lachesis row: now "Design range 150–400 …; hard cap 400", cross-referencing the artifact.
• V6/evals/cost-budget/launch-token-budgets.json Lachesis concurrency cap: values unchanged (designTarget 400 / hardCap 400 — consistent with the adopted story); added a note field stating designTarget 400 is the upper point of the 150–400 design range.

4. Launch DAU projection quantified (was load-bearing but never defined)#

• V6/V6_DEPENDENCIES.md §19: added canonical block "Launch DAU projection (canonical planning assumption, adopted 2026-06-12)" — 30,000 DAU (assumption, not a market forecast) → ≈10,000 DAU/region across the 3 sustained-load regions (iad/fra/sin) → 10% peak concurrency ≈1,000 concurrent players/region → ≈10.2 agents per concurrently-active player (≈10 Lachesis from the "company you can know" household scale, features§"The Household Roster", + 0.2 Clotho ≈ 13 full ≤16-agent active scenes per region, matching the manifest's clothoActiveSceneCount: 13) → ≈10,200 concurrently-resident agents per region = the §42 fixed mix (10,000 Lachesis + 200 Clotho). Also states the equivalent decomposition (≈25–66 fully-resident Solo worlds at 150–400 each) and that a real forecast above 30,000 DAU forces re-derivation.
• Gate references now point at the number:
  • V6/V6_features.md §"Launch Readiness" Moirai bullet,
  • V6/V6_ARCHITECTURE.md §"Launch Readiness" Moirai bullet,
  • V6/V6_TODOS.md "V6 Exit Criteria (high level)" bullet and §48 Moirai item (the §48 item also re-marked [~], see §9 below).

5. Eval-set taxonomy mismatch (5 named vs 13 required)#

• V6/V6_TODOS.md §41 first task (text only, checkbox kept [x]): now enumerates all 13 suites V6/evals/agent-behavior/suites.json requires (value-refusal, continuity, grounding, learning-by-example, negotiation, objective-assignment, goal-arc, gifts-opportunity, life-arc, crossroads, persona-policy, crisis, minor-protection), noting persona-policy / crisis / minor-protection are safety-classified per features§"Safety Evals".
• V6/V6_DEPENDENCIES.md §35: "Agent-behavior eval harness" row enumerates the 13 suites with the safety-classification note; "Safety eval harness" row cross-references that the same three suites are the safety-classified ones.
• V6/V6_features.md:1720-1728 safety classification left intact (as required).

6. Commons scale claim vs test (thousands promised, 32 measured)#

• V6/V6_features.md §"The Commons" "shared world" bullet: now states the AoI/interest-management design explicitly — 4,096 agents simulated, per-client fully-replicated visible set capped ~32, up to ~150 drawn via density LOD with crowd impostors/silhouettes beyond, bandwidth measured against the capped visible set.
• V6/performance/commons-scale-load.v6perf.json: added measurementScopeNote field clarifying what was measured (bandwidth at visibleAgentCount 32, not replication of thousands to one client) and that console/mobile platformMeasurements are repo-authored projections, not captured device runs.

7. Launch locales + accessibility standard never named#

• V6/V6_features.md §"Localization": added a Launch locales paragraph — exactly en-US, es-ES, yo-NG, citing V6/localization/region-rating-residency.v6loc.json (checkedLocales).
• V6/V6_features.md §"Accessibility": named the standard — WCAG 2.2 Level AA, evaluated by axe-core in CI plus per-platform a11y suites for the UE client.
• V6/V6_TODOS.md §39 fourth task: acceptance text now reads "green against WCAG 2.2 AA …" (text only; checkbox unchanged).

8. V2–V5 liveness dependency re-scoped (coherence break)#

• V6/V6_features.md §"Launch Readiness" Aye bullet: "All four Aye Thresholds operational … for every realm that is itself live" → integration-tested against per-destination adapters and realm stubs; production liveness of each destination realm is an EXTERNAL dependency tracked outside V6 and re-verified at each realm's GA (none of V2–V5 is live as of 2026-06-12).
• V6/V6_DEPENDENCIES.md Aye Threshold readiness gate paragraph: same re-scope; the gate explicitly does not prove V2–V5 are operational live realms.
• Checkbox re-marks (liveness-dependent):
  • §48 "All four Aye Thresholds operational with round-trip journals for every live Aye realm" — [x]→[~] (destination-realm liveness external).
  • §38 "[P2] Implement Aye campaigns" — [x]→[~] (an open V2 season / V5 chapter requires a live destination realm).
  • §27 Aye Bridge implementation tasks left [x] — they are locally verifiable against adapters/stubs and do not claim realm liveness.

9. Externally-unverifiable evidence marked done#

• (a) Platform cert: V6/release/CROSS_PLATFORM_CERTIFICATION.md — provenance caveat added above the Required Outcomes table ("approved and published" rows are pending vendor evidence; approvals cannot be proven from repo artifacts; the verifier does not call store APIs). V6/V6_TODOS.md §45 "Pass platform cert …" — [x]→[~].
• (b) Staging telemetry: V6/evals/cost-budget/staging-token-telemetry.json — added provenance: "synthetic-fixture" and provenanceNote (hand-authored fixture; no V6 staging environment has emitted real Moirai telemetry; real telemetry pending).
• (c) Other real-world-operation claims in TODOS, each judged individually:
  • §42 "Run a sustained load test … 7 consecutive days" — [x]→[~] "structural evidence only; operational run pending" (the manifest's evidence is local cargo tests that complete in seconds; a 7-day 3-region infra run cannot have occurred). Matching provenanceNote added to V6/performance/sustained-load.v6perf.json.
  • §42 "Run a Commons-scale load test" — [x]→[~] (real local bandwidth/cognition tests, but console/mobile frame/memory figures are repo-authored projections; see artifact measurementScopeNote).
  • §40 "Implement region rating variants …. Done when: shipped locales meet their region rating" — [x]→[~] (ESRB/PEGI/IARC outcomes in the locale artifact are external rating-board decisions for an unshipped product; the variant mechanism itself is repo-verifiable).
  • §48 Moirai launch-readiness item — [x]→[~] (depends on the §42 sustained-load operational run and the synthetic §36 staging telemetry).
  • §48 "Accessibility, localization, cert, security …" aggregate — [x]→[~] (its Done-when requires §40/§45 items now [~]).
  • §48 "V6 Exit Criteria all met" — [x]→[ ] (contradicted: unchecked and [~] tasks now exist in the document).
  • Judged and left [x]: §14 "≥150 grounded wild agents seeded" (seed pool is a repo fixture), §47 runbooks/production-setup (docs exist in repo), §45 cross-progression (fixture-verifiable), §44 DSAR (code-verifiable).

10. District "authoring" honesty#

• V6/V6_ARCHITECTURE.md §"Game Feature Plugins": added a District scene descriptors — disclosure paragraph defining the descriptor convention (DistrictScene_<Name>.v6district.json = design artifact / machine-checked contract, not cooked UE content; V6/ue/Content/ is .gitkeep-empty; in-engine authoring requires World Partition maps, meshes, lighting, navmesh, cooked content).
• V6/V6_TODOS.md §28, judged per task wording:
  • All six "Author the " tasks (Grove of Beginnings, Ateliers, Commons Heart, Wilds, Ancestor Grove, Thresholds) — [x]→[ ] "descriptor only; no UE content exists" (each cites its descriptor file).
  • §28 "Sign every district scene … and promote" — left [x]: it operates on the scene descriptors (signing/promotion of the artifacts), not on in-engine world authoring.
  • §48 "All six districts of Orun open and authored" — [x]→[ ] (contradicted by the §28 reverts).

11. Vague-in-doc, quantified-only-in-artifact#

• V6/V6_features.md §"Performance Budgets": Chronicle "within a few seconds" now pinned ≤ 3,500 ms; shard/Threshold "perceptible stall" now pinned ≤ 50 ms max client stall; escalation-rehydration grace window DEFINED: ≤ 500 ms marked "budget adopted 2026-06-12; measurement pending".
• V6/V6_ARCHITECTURE.md §"Performance Budgets": same three numbers added.
• V6/performance/latency-budgets.v6perf.json: added sharedBudgets.escalationRehydrationGraceMs: 500 plus a budgetNotes entry stating it is a target only with no observed measurement (no assertion claims one) — honestly recorded as adopted-not-measured.

12. Completion-state banner#

• V6/V6_TODOS.md: warning banner added near the top (modeled on V5/V5_TODOS.md:40-54): V6 has never received an adversarial implementation verification; [x] attests to repo artifacts only, not operational proof or vendor approvals; operationally-unverifiable items re-marked [~] on 2026-06-12 per this log.

Checkbox re-mark summary#

Known consequence: verifier pins will fail closed (intentional)#

Several scripts/v6/verify-v6-*.mjs gates pin the exact former - [x] ... TODO lines and will now fail (fail-closed), including at least: verify-v6-cross-platform-certification.mjs, verify-v6-sustained-load.mjs, verify-v6-commons-scale-load.mjs, verify-v6-region-rating-residency.mjs, verify-v6-aye-campaigns.mjs, verify-v6-aye-threshold-readiness.mjs, verify-v6-moirai-cost-load-readiness.mjs, verify-v6-grove-of-beginnings.mjs (and the other district verifiers), verify-v6-accessibility-localization-security-cert-readiness.mjs, and verify-v6-exit-criteria-readiness (which requires zero unchecked boxes). Post-edit spot-runs: run-cost-budget-harness.mjs and verify-v6-latency-budgets.mjs still pass; verify-v6-sustained-load.mjs and verify-v6-commons-scale-load.mjs now fail closed on the re-marked checkboxes as predicted (verify-v6-docs-drift.mjs fails for a pre-existing, unrelated reason: the VRM4U submodule is not initialized in this checkout). This is the honest outcome: those gates were attesting to claims the repo cannot prove. Updating the gate scripts to reflect the re-scoped claims is follow-up engineering work, deliberately not done here (this remediation does not edit code outside V6/).

All five touched JSON artifacts re-parse cleanly (latency-budgets.v6perf.json, sustained-load.v6perf.json, commons-scale-load.v6perf.json, launch-token-budgets.json, staging-token-telemetry.json); the fixed anchor slug performance-budgets matches the real ## Performance Budgets heading.