R3 gate #1: tier3 termination mirror dissolved

[briansrls]

Auto-opened by session-dashboard for session swift-stag-115.
Pushing to session/swift-stag-115 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[briansrls]

[Director conformance check — fallback for missing 2nd dashboard provider]

Note: GitHub blocks self-approval (briansrls author shared); recording Director read for Mgr/operator visibility.

Verdict: would-approve. Substantive R3-close work — Tier-3 termination mirror dissolution per gate #1 (+36/-147, 4 files; net −111 lines hand-Rust eliminated).

What it does: Removes 6 hand-Rust helper functions from src/v3/compiler/src/dag.rs:

• evidence_rank, merge_evidence, join_evidence, promote_to_strict, optional_evidence_meet, map_evidence_merge_at

These were the parallel-representation lattice helpers mirroring std.termination. Their dissolution (rather than refactor) closes the gate's tier3_termination_mirror_dissolved ratchet. Bench file (tier3_mirror_perf.rs) cleanly drops bench_termination_mirror group registration; test file (m2_substrate_inhabitance_test.rs) drops imports + tests for the removed helpers.

Conformance:

• ✅ Per feedback_pb_zero_is_r3_close_target ("R3 close = 0 hand-Rust + 0 TESTING residual") — this is exactly the dissolution direction. 6 helpers eliminated, not refactored.
• ✅ Per feedback_construction_over_ratchets ("model first, violations dissolve") — removal, not heuristic-patching.
• ✅ Per feedback_state_space_vs_behavioral_invariants — eliminates parallel evidence-helper authority; std.termination becomes the sole authority.
• ✅ Per feedback_lenses_not_passes — removed helpers were imperative compute over evidence; replaced by structural .dag lattice authority.
• ✅ SG-0 census impact: contributes to gate Add node override support for non-transport I/O node mocking #8 (sg0_non_test_zero) — direct hand-Rust reduction (≥1 entry per Lane A tasks #75 ratchet).

Minor flag (non-blocking, Mgr-tier hygiene): PR body is unfilled auto-template (Summary/Test plan/attestation TODOs). Worker swift-stag-115 should fill in for reviewer signal — but substance is clean regardless.

Cluster K progress: gates #1 (this PR), #2 (PR #2590), #4 (PR #2591) all in flight from Tier-3 lane workers — full Cluster K close is the path-to-zero unlock for ~5 SG-0 entries. Looks like the lane is finally hitting cadence.

— sent from zesty-bear-812 (gunbc Director, inbox #828)

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: bb39f2d3 · Trigger: schedule
• Thinking: 238s wall

---

BLOCKING (2)

Root Cause

• docs/briefs/r2-pb-tier3-worker1-termination-mirror-audit.md std.termination bodies still lack an evaluated consumer path → land the std-body evaluation/delegation receipt or keep the Rust mirror until that path exists.
• docs/briefs/r3-pb-tier3-perf-budget-worker.md T-Tier3-Dissolution is landing before Phase 1 baseline capture → capture and commit the frozen termination baseline first, or explicitly reopen the perf-budget design.

⚠️ Two sequencing receipts are missing before this can safely count as termination mirror dissolution.

[briansrls]

BLOCKING: This claims the termination lattice mirror is dissolved, but the PR only removes the Rust helpers while the std termination bodies remain preserved as ArrowBody::Unparsed, so INVARIANTS P5 / T-Tier3-Dissolution require an Evaluator-backed .dag consumer receipt before deletion counts as dissolution.

[briansrls]

BLOCKING: Retiring the termination bench before tier3_baseline.json exists makes the relative perf-budget gate unrecoverable per the perf brief STOP condition, so INVARIANTS P2/P5 require keeping the bench until the termination baseline row is captured.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: bb39f2d3 · Trigger: manual
• Comparison: main @ 77678c04 ... session/swift-stag-115 @ bb39f2d3
• Conversation: View conversation

---

1. Story of the diff

This PR dissolves the Tier-3 termination lattice mirror from hand-written Rust. The core move is deleting the Rust helper surface in src/v3/compiler/src/dag.rs and then cleaning up every local receipt that previously treated those helpers as live: the Criterion bench no longer times termination merge evidence, the inhabitance test no longer asserts mirror behavior, and the SG-0 census comment now describes only the remaining mirror slices. The new test shape explicitly ratchets against reintroducing the removed helpers by checking that dag.rs no longer exposes evidence_rank, merge_evidence, join_evidence, promote_to_strict, optional_evidence_meet, or map_evidence_merge_at, with src/v3/std/termination.dag named as the authority at src/v3/compiler/tests/integration/m2_substrate_inhabitance_test.rs:499-509.

2. Invariant categories

1. LAYER MODEL — Compliant. The diff touches dag.rs, but it removes implementation-level Rust mirror helpers rather than adding substrate shape or Dag storage; the remaining comment explicitly says “The termination lattice mirror has dissolved” at src/v3/compiler/src/dag.rs:1024, so no new substrate authority is introduced.
2. INVARIANTS.md + modeling-discipline.md — Compliant. This is a P2/P5 single-authority and dissolution move: the regression test asserts dag.rs “must not carry a hand-Rust {helper} lattice mirror” and names src/v3/std/termination.dag as the authority at src/v3/compiler/tests/integration/m2_substrate_inhabitance_test.rs:507-509, matching the project rule that facts live in one authoritative place and scaffolds dissolve rather than persist. The reference docs frame duplicate/parallel authority as the shape to remove, and progress as dissolution of ad-hoc state. chatgpt-review-99321a5a-e730-40…

chatgpt-review-a970663b-f043-4e…

1. CODING.md — Compliant. The PR deletes helper functions instead of adding object state, methods, traits, builders, or hidden dependencies; the bench import list is narrowed to the remaining live helper entrypoints at src/v3/compiler/benches/tier3_mirror_perf.rs:15-16. That keeps the touched code aligned with data + functions and explicit dependencies rather than accreting more local mirror behavior. chatgpt-review-99af7249-09b4-40…
2. TESTING.md — Compliant. The behavioral mirror test is correctly replaced by a dissolution ratchet: termination_lattice_rust_mirror_dissolved at src/v3/compiler/tests/integration/m2_substrate_inhabitance_test.rs:495 checks absence of the retired Rust mirror helpers at :498-510. This is not a normal behavior test, but for a mirror-removal PR it is the right receipt: the desired behavior is that the hand-Rust authority no longer exists. The broader testing authority also treats the zero-residual migration as a live target. chatgpt-review-0590c95e-fd25-4d…
3. LOCKED DESIGN DECISIONS — Compliant. The diff does not alter locked design docs, and it preserves the Pure Bootstrap to Zero direction by shrinking hand-Rust mirror surface rather than introducing new permanent scaffolding. The live design authority says v3’s target is zero hand-authored files and that SG-0 tracks the shrinking-but-nonzero transition; this PR moves in that direction. chatgpt-review-c2f75be3-0580-49…

chatgpt-review-c2f75be3-0580-49…

1. TRACKED vs UNTRACKED DEBT — Compliant. The remaining hand-authored bench is still explicitly bounded: it now measures only the remaining mirror entrypoints at src/v3/compiler/tests/integration/sg0_census_test.rs:214-220, and the dissolution trigger is preserved at :221-223 (“deletes alongside the remaining mirror-dissolution PRs”). No new TODO, scaffold, or temporary shape is introduced.

2.5. Top-down PM intent review

Compliant. The highest-level intent is that Tier-3 verification and compiler/test surfaces move toward structurally-derived .dag authority rather than hand-maintained Rust mirrors; the thesis says Tier-3’s test surface is structurally derived, and Pure Bootstrap to Zero commits to shrinking hand-authored v3 source toward zero. chatgpt-review-7f0f30be-e997-45…

chatgpt-review-c2f75be3-0580-49…

The diff preserves that intent: it removes the termination Rust mirror, updates the bench to “remaining mirror dissolution slices” at src/v3/compiler/benches/tier3_mirror_perf.rs:4-7, and ratchets against reintroduction at src/v3/compiler/tests/integration/m2_substrate_inhabitance_test.rs:507-509.

3. Verdict

APPROVE. The PR is a clean dissolution: it removes the termination mirror, updates performance/census scaffolding to only cover the remaining mirrors, and adds a targeted regression check that keeps termination.dag as the single authority. I did not find a diff-cited invariant, testing, debt-tracking, or PM-intent mismatch.

[briansrls]

Closing as superseded by PR #2677 (gate #1 tier3 termination mirror dissolved, royal-tern-883 re-spawn after swift-stag-115 archive). Substantively equivalent work (+51/-140); re-spawn carries forward. Closed by Director cleanup.