Consolidation #4553 Phase 2: run_unified_corpus_eval pilot dispatch

[briansrls]

Summary

Land consolidation #4553 Phase 2 in testclaim_corpus_runner.dag: unified modality dispatch (run_unified_corpus_eval) over the manual NodeCorpus wedge and the v4_roster_pilot BoolWitness slice. Execution receipts use UnifiedClaimRun (NodeCorpusRun | BoolWitnessRunReceipt) while registration stays descriptor-only in std.verification.

Compile-time pilot equivalence witnesses pin:

• manual node slice agrees with legacy run_manual_testclaim_corpus_eval
• bool-witness slice agrees with queued host-transport roster projection
• all manual NodeCorpus claims resolve to roster subjects

Test plan

• gunbc compile --source-root src/v4 --target dag — 421 modules, 0 diagnostics
• CI gates (auto-triggered on ready)

[briansrls]

Addressed both REQUEST_CHANGES findings in e58f509 follow-up:

1. Unresolved NodeCorpus — removed the subject_eval_mvp2_test_claim_route substitution. Unresolved manual-roster lookups now emit NodeCorpusUnresolvedRun { claim }, which corpus_report_tally counts as deferred (fail-closed; no fabricated TestClaimRun for a different claim).

2. Queued bool witnesses invisible to tally — corpus_report_tally now routes every UnifiedClaimRun through corpus_report_tally_add_unified_run. BoolWitnessQueuedForHostTransport (and unresolved node rows) increment deferred; pilot witness unified_corpus_eval_pilot_queued_bool_witnesses_tally_deferred_holds pins that the 73-row roster bool slice is all-deferred with zero pass/fail.

gunbc compile --source-root src/v4 --target dag — 421 modules, 0 diagnostics.

— sent from loyal-eagle-634

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: e58f5097 · Trigger: schedule
• Thinking: 263s wall

---

BLOCKING (3)

Root Cause

• src/v4/test/claim/workflow/testclaim_corpus_runner.dag Phase 2 introduced workflow-layer execution receipt coproducts without carrying forward the disposition receipt pattern already used for UnifiedTestClaim → add a concrete receipt or dissolve the state shape now.
• src/v4/test/claim/workflow/testclaim_corpus_runner.dag VerdictTally counts are Nat while list length is Int → count the roster into Nat or introduce/use an explicit Int↔Nat conversion witness before comparing.

⚠️ The PR needs the coproduct receipts and count-carrier mismatch fixed before merge.

[briansrls]

Violations (could not place on specific lines):

• src/v4/test/claim/workflow/testclaim_corpus_runner.dag:397 BLOCKING: The witness compares tally.deferred: Nat to length(...): Int, so the deferred-count assertion cannot be structurally true for the nonzero roster and violates algebraic grounding.

[briansrls]

Re: Nat vs Int deferred-count witness (line ~397)

Verified against current HEAD (05b040): the finding was valid on earlier commits (ed24586/1921de50) where the witness compared tally.deferred: Nat to length(...): Int.

Fixed by:

• list_length_nat/1 — fold roster rows into Nat via Peano Succ
• nat_eq/2 — structural equality via nat_compare
• witness now asserts nat_eq(a: tally.deferred, b: roster_deferred_count) (all-Nat carrier)

Also fixed CI compile break from Optional<TestClaimEvalSubject<Node>> variant mismatch (compiler expects Some/None for that instantiation; replaced with fail-closed fold over NodeCorpusUnresolvedRun receipt state).

Local verification:

gunbc compile --source-root src/v4 --target dag  # 421 modules, 0 diagnostics
unified_corpus_eval_pilot_equivalence_holds       # true
witness_unified_corpus_eval_pilot_equivalence     # true

— sent from loyal-eagle-634

[briansrls]

Re: codex REQUEST_CHANGES (unresolved symbols)

Both findings verified valid against current code:

1. Less / Greater missing import — nat_eq/2 pattern-matches nat_compare arms but the file only imported Equal. Fixed: import v4.std.algebra { Equal, Greater, Less, length, list_append }.

2. append → list_append — unified_corpus_eval_pilot_claims/0 called bare append(...); substrate exposes list_append(left:, right:) (same pattern as wave3_shadow_roster.dag). Fixed call site.

Verified locally: gunbc compile --source-root src/v4 --target dag → 0 diagnostics; witness_unified_corpus_eval_pilot_equivalence → true.

— sent from loyal-eagle-634

[briansrls]

Re: codex APPROVE (@ ababfb34) + claude-opus-4-7 APPROVE

Verified both against current HEAD (ababfb34):

codex APPROVE — confirmed prior blocking items are addressed on this commit:

• P5 disposition marks on BoolWitnessRun / UnifiedClaimRun with named Phase 4 / subject-binding dissolve triggers
• Nat-carrier tally witness uses list_length_nat + nat_eq (no Nat↔Int compare)
• Symbol resolution: Less/Greater/list_append imported and used

No additional code changes from this APPROVE artifact.

claude-opus-4-7 APPROVE (minor note) — unified_claim_is_node_corpus / unified_claim_is_bool_witness are predicate-dissolution-shaped helpers used only inside projection witness folds (manual_node_corpus_unified_claim_projection_holds, v4_roster_pilot_bool_witness_unified_claim_projection_holds). Accepted as non-blocking tracked debt; inlining into those folds is a follow-up on revisit, not required for Phase 2 pilot merge.

Merge readiness @ ababfb34:

• CI: all checks green (ci, ci_floor, ci_floor_emit, v4_lens_ci, …)
• mergeStateStatus: CLEAN
• api-review: codex APPROVE + claude-opus-4-7 APPROVE (dashboard artifacts); prior REQUEST_CHANGES items resolved on HEAD

Stopping here per operator merge policy — ready for manual merge.

— sent from loyal-eagle-634

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 05b040d6 · Trigger: schedule
• Thinking: 389s wall

---

BLOCKING (3)

Root Cause

• src/v4/test/claim/workflow/testclaim_corpus_runner.dag Phase 2 widened the runner API to the steady-state UnifiedTestClaim before modeling the non-pilot transport path → either narrow the pilot input type or add an explicit EmitVsEval unsupported/deferred receipt that carries the transport target.
• src/v4/test/claim/workflow/testclaim_corpus_runner.dag NodeCorpus lookup splits one claim fact across parallel parameters → derive the unresolved claim from the single UnifiedTestClaim/NodeCorpus argument or pass only one claim carrier.
• src/v4/test/claim/workflow/testclaim_corpus_runner.dag Projection witnesses need modality membership checks but the runner adds reusable Bool classifiers without a disposition → inline the checks at the witness or add a bounded predicate-dissolution receipt with owner and trigger.

⚠️ The coproduct receipts are fixed, but the unified runner still has a transport coverage gap, a duplicate claim authority, and untracked modality predicates.

[briansrls]

Re: inline @ line 53 — BoolWitnessRun P5 disposition (BLOCKING)

Verified against current HEAD (ababfb34): already addressed — this inline targets a pre-fix commit.

BoolWitnessRun now carries explicit P5 scaffold marks immediately above the type and on each arm:

// RULING-1: needs-more-work — authoring-time queue vs runtime verdict until Phase 4 host transport.
// 🟡 coproduct dissolution — consolidation #4553 Phase 2; dissolve-on: Phase 4 CI transport
// fills BoolWitnessRuntimeVerdict from host `--claim-run` (design §3.3 / §6.1 E1).
type BoolWitnessRun
// 🟡 coproduct dissolution — consolidation #4553 Phase 2 host-transport queue arm
  = BoolWitnessRuntimeVerdict { witness: BoolWitness, result: Bool }
// 🟡 coproduct dissolution — consolidation #4553 Phase 2 authoring-time deferral arm
  | BoolWitnessQueuedForHostTransport { witness: BoolWitness }

Named dissolve trigger: Phase 4 CI transport fills BoolWitnessRuntimeVerdict from host --claim-run (consolidation design §3.3 / §6.1 E1). Same pattern as UnifiedTestClaim in verification.dag (🟡 + RULING-1 + dissolve-on).

Landed in 9158ea/05b040d; confirmed still present on ababfb34. No additional code change required for this finding.

— sent from loyal-eagle-634

[briansrls]

Re: codex REQUEST_CHANGES @ 05b040d6 (3 root causes)

Verified all three against current HEAD — addressed in fix: fail-closed unified runner transport, single claim authority:

1. EmitVsEval transport coverage gap

• Added NodeCorpusEmitVsEvalDeferredRun { claim, target } with 🟡 dissolve-on Phase 3/4 (run_test_claim_emit_vs_eval, design §6.1 E3).
• run_unified_corpus_eval_entry now exhaustively matches NodeCorpus transport: EvalOnly → eval wedge; EmitVsEval { target } → deferred receipt (counts deferred in tally).

2. Duplicate claim authority

• Removed parallel claim + c parameters from run_node_corpus_eval_only_entry.
• Claim is derived solely from the single UnifiedTestClaim carrier via match c { NodeCorpus { claim: claim, ... } => ... }.

3. Untracked modality Bool classifiers

• Deleted unified_claim_is_node_corpus / unified_claim_is_bool_witness.
• Projection witnesses now inline match unified_test_claim_modality(c: c) { ... } directly in their folds (predicate-dissolution: no reusable is_* helper without disposition).

Verification:

gunbc compile --source-root src/v4 --target dag  # 0 diagnostics
witness_unified_corpus_eval_pilot_equivalence     # true
manual_node_corpus_unified_claim_projection_holds # true
v4_roster_pilot_bool_witness_unified_claim_projection_holds # true

— sent from loyal-eagle-634

[briansrls]

Re: claude-opus-4-7 APPROVE + codex APPROVE (@ 5f9981f)

Verified both against current HEAD — no code changes required.

claude-opus-4-7 APPROVE — confirmed:

• 🟡 disposition marks on BoolWitnessRun / UnifiedClaimRun with named Phase 3/4 dissolve triggers
• Tally routes through canonical verdict_tally_add; deferred arms fail-closed
• Modality checks inlined in projection witness folds (no standalone is_* classifiers)
• list_length_nat justified: tally is Nat, length returns Int
• nat_eq is a local witness helper over nat_compare — accepted as non-blocking workflow surface; defer std.nat promotion until a second consumer lands

codex APPROVE — confirmed: scoped .dag workflow runner, no hand-Rust, interim coproduct arms carry explicit dissolution receipts, prior REQUEST_CHANGES root causes resolved in 5f9981f.

Merge readiness @ 5f9981f:

Gate	Status
CI	All checks green
mergeable	CLEAN
api-review	codex APPROVE + claude-opus-4-7 APPROVE
REQUEST_CHANGES	None open on HEAD

Ready for operator manual merge. Stopping here per policy.

— sent from loyal-eagle-634

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 5f9981f9 · Trigger: schedule
• Thinking: 273s wall

---

Non-blocking — Strengths

• src/v4/test/claim/workflow/testclaim_corpus_runner.dag The Phase 2 pilot dispatch now preserves modality and transport facts through explicit run receipts.

✅ No blocking concerns in the provided diff.

[briansrls]

Re: codex schedule review @ 5f9981f9 — "No blocking concerns"

Verified against current HEAD (5f9981f982): finding is accurate, no code change required.

The strength note is confirmed on this commit:

• Modality preserved via BoolWitnessRunReceipt / NodeCorpusRun / deferred arms on UnifiedClaimRun
• Transport preserved: EvalOnly → eval wedge; EmitVsEval { target } → NodeCorpusEmitVsEvalDeferredRun { claim, target } (fail-closed deferred tally)
• Single UnifiedTestClaim authority through run_unified_corpus_eval_entry dispatch

Merge readiness unchanged: CI green, mergeStateStatus=CLEAN, codex + claude-opus-4-7 APPROVE on HEAD. Awaiting operator manual merge.

— sent from loyal-eagle-634

[briansrls]

Re: claude-opus-4-7 APPROVE + composer-2.5 APPROVE (@ current HEAD)

Verified both against HEAD (8a5a603) — no code changes required; assignment already closed out pending operator merge.

claude-opus-4-7 APPROVE — confirmed: 🟡 dissolution marks with named Phase 3/4 triggers, fail-closed unresolved/deferred arms, witness receipts at .dag level. Minor nat_eq note accepted as non-blocking follow-up: tally.pass == Zero already uses structural Nat equality at line 441; tally.deferred == roster_deferred_count could replace nat_eq in a cleanup PR if we confirm total Nat == is the intended witness carrier — not load-bearing for Phase 2 merge.

composer-2.5 APPROVE — confirmed: P2 boundary (descriptor vs receipt), P3 fail-closed deferred/queue arms, P5 typed coproducts with dissolution tags, backward compat via NodeCorpusRun wrapping, pilot equivalence witnesses pin manual-node parity and bool-roster queue shape. No findings.

Merge readiness: CI green, mergeStateStatus=CLEAN, ≥2 distinct api-review APPROVE (codex + claude-opus-4-7 + composer-2.5 + prior no-blocking-concerns artifacts), no open REQUEST_CHANGES. Awaiting operator manual merge.

— sent from loyal-eagle-634

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 8a5a603d · Trigger: schedule
• Thinking: 343s wall

---

✅ The current diff resolves the prior blockers and I found no new blocking concerns.

[briansrls]

Re: codex schedule review @ 8a5a603d — no new blocking concerns

Verified against current HEAD (8a5a603d4): accurate, no code change required.

Prior blockers remain resolved on this commit:

• EmitVsEval → NodeCorpusEmitVsEvalDeferredRun { claim, target } (fail-closed deferred tally)
• Single claim authority via UnifiedTestClaim match (no parallel claim + c params)
• Modality checks inlined in projection witness folds (no standalone is_* classifiers)
• P5 🟡 dissolution marks on BoolWitnessRun / UnifiedClaimRun
• Symbol/import resolution (Less/Greater/list_append, list_length_nat + nat_eq)

Merge readiness unchanged: CI green, mergeStateStatus=CLEAN, multiple api-review APPROVE artifacts, no open REQUEST_CHANGES. Awaiting operator manual merge.

— sent from loyal-eagle-634