:Lens<EmissionProvenance> carrier + structural lens shape

[briansrls]

Summary

Per Director Q1(a) RATIFIED at gunbc#1739 #issuecomment-4392562911 — emission-provenance is a per-Behavior Lens<C>-shaped analysis (rejecting Q1(b) per-line instrumentation and Q1(d) parallel substrates).

Brief: docs/briefs/r3-substrate-emission-provenance-lens-worker.md.

What landed

• src/v3/std/emission_provenance.dag — typed carrier

  • EmissionRule = String (field-path; per Director Q3 feedback_reason_not_label)
  • OptionalSourceSpan = NoSourceSpan | SomeSourceSpan { value: SourceSpan } (typed-sum idiom; v3 std has no generic Option<T>)
  • EmissionProvenance { emitted_line, rule, source_span } — record (product) per codex BLOCKING reshape (PR R3 Substrate #1910 sha 2ed1046 Finding Add SVG viz, test helpers, and makegen scaffold #1)

• src/v3/lenses/emission_provenance.dag — Lens<List<EmissionProvenance>> structural fns (read_provenance / concat_provenance / branch_provenance / iterate_provenance / validate_provenance) per Director-locked 6-field Lens<C> shape

• Cementing tests (emission_provenance_lens_test.rs):

  • emission_provenance_record_has_locked_three_field_shape — passes
  • optional_source_span_carries_none_and_some_arms — passes
  • emission_provenance_lens_binds_against_locked_carrier — #[ignore] (substrate-cascade STOP, see below)

Substrate-cascade STOP

Promoting data emission_provenance_lens: Lens<List<EmissionProvenance>> to a top-level declaration is blocked by Class 5 Gap 3 (top-level ValueBody boundary; sum-variant literals like Empty don't lower in data body context). Same constraint blocks the canonical generic data list_monoid<element>: Monoid<List<element>> documented at src/v3/std/list.dag:61-64.

Worker followed the named_function_count.dag / dag_shape.dag precedent (lens .dag files declare structural fns + types; data <name>_lens: Lens<C> instance lives in fixture context per e6_g1a_option3_static_lens_test.rs::mini_lens).

The #[ignore]d test contains the full fixture-bound lens declaration and serves as the dissolution-trigger landing pad: when Class 5 Gap 3 closes, un-ignore the test and (separately) promote the data declaration to the top-level lens .dag without lens-shape changes.

What's deferred

BEHAVIORALLY DEFERRED — emitter writes field-path → lens recovers same field-path round-trip. Producer wiring (instrumenting render_named_template call sites in rust_target.rs) is a separate downstream slice. Precedent: lenses.cost BEHAVIORALLY PROXY, lenses.dag_shape producer-only.

NOT bundled per Director ratification at gunbc#1739 #issuecomment-4392225548:

• T-Rule-Enumeration edits (RETIRED per Reading C ratification)
• Grounding-side visualization-consumer wiring

5-question authority audit

1. Substrate exists? EmissionProvenance carrier + lens fns — this PR is producer. Lens<C> carrier landed (src/v3/std/lens.dag, 🟢 TERMINAL). T-Rule-Enumeration retired (Director Reading C — field-path enumeration is structurally already provided by *SyntaxBinding / *OpsBinding field set; substrate-state-grep confirmed at gunbc#1759 #issuecomment-4392696623).
2. Existing brief? r3-substrate-emission-provenance-lens-worker.md (parent canvas: r3-substrate-emission-provenance-shape-canvas.md). PM-authored proposal at PR docs(briefs): Lens<EmissionProvenance> worker brief — PM-authored, Director-ratified #1902 superseded by Substrate Mgr canvas + Director Q1(a) re-ratification.
3. Design-doc match? Director Q1(a) RATIFIED + canvas Q1 disposition + Director-locked Lens<C> 6-field shape. T-CostLens-Composition is shape precedent.
4. Citations live? Director Q1+Q2+Q3 RATIFICATION at gunbc#1739 #issuecomment-4392562911; Reading C RATIFICATION at #issuecomment-4392797954; codex BLOCKING reshape at PR R3 Substrate #1910 sha 2ed1046 Finding Add SVG viz, test helpers, and makegen scaffold #1.
5. Carrier dissolves the bridge? Yes for the carrier. Record EmissionProvenance with mandatory rule + optional source_span dissolves "what produced this emitted line?" structurally (rule field always present; type system enforces). Lens-instance promotion to top-level data cascades on Class 5 Gap 3.

Test plan

• cargo build --workspace clean
• cargo fmt --all --check clean
• cargo clippy --all-targets -- -D warnings clean
• new tests pass (2 of 3; 1 #[ignore] per substrate cascade above)
• cargo test --workspace --exclude v2-compiler-tests green (CI verifies)
• cargo test -p v2-compiler-tests green; strict-compile ratchet at 0 (CI verifies)

SG-0 net-shrink discipline

SG-0 hand-path delta: +1

SG-0 pairing: (c) structural deferral + named follow-up dispatch — the new test path src/v3/compiler/tests/integration/emission_provenance_lens_test.rs is the dissolution-trigger landing pad for the fixture-bound lens instance. Named follow-up: when Class 5 Gap 3 closes (top-level ValueBody carriers + fn-typed data-body fields land), un-ignore the emission_provenance_lens_binds_against_locked_carrier test and (separately) promote the data emission_provenance_lens declaration to top-level in src/v3/lenses/emission_provenance.dag. ROADMAP "Class 5 Gap 3" row tracks the dissolution.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: c2cd8495 · Trigger: schedule
• Comparison: origin/main @ 63044050 ... review/pr-1928-c2cd8495 @ c2cd8495
• Thinking: 27s wall

---

Findings: none blocking.

Verdict: APPROVE — Small, additive substrate slice. New EmissionProvenance carrier is record-shaped with mandatory rule and OptionalSourceSpan typed-sum (no Option<T>), matches v3 std idiom, and is structurally fail-closed (illegal-state-unrepresentable for "rule absent"). Lens fns are pure, data + free functions per CODING.md. The #[ignore]'d cementing test has all three properties of tracked debt: documented, bounded, named dissolution trigger (Class 5 Gap 3 closure → un-ignore). The read_provenance stub returning Inhabits(Empty) is honest fail-closed scaffolding tied to a separately-tracked downstream slice (render_named_template instrumentation), not silent fabrication.

Exploratory observations:

• The two doc blocks state the Class 5 Gap 3 blocker slightly differently — the file header at lenses/emission_provenance.dag:103 says "sum-variant literals like Empty don't lower in data body context," while the test attribute at emission_provenance_lens_test.rs:184 cites "opaque-body lowering rejects the nested-fn-reference identity empty_provenance_list in monoid identity field." Both plausibly true, but if both are real distinct blockers, the dissolution trigger ("just remove #[ignore]") may be optimistic — closing one gap may not unblock the other. Worth a sentence reconciling them when the gap closes.
• EmissionRule = String is pragmatic per Director Q3 ratification, but it's the classic "stringly-typed enum" — every consumer will need to defend against malformed paths. Not a blocker (the doc is explicit this is by-design and the field-path enumeration is structurally pinned by the binding structs), just worth noting that the validate_provenance lens body will eventually need to do that defense, since the type can't.

[briansrls]

Addressed both claude exploratory observations (sha c2cd849) in 32302cb:

1. Reconciled two Class 5 Gap 3 phrasings as facets of the same root (data bodies don't accept top-level non-scalar/non-record carriers). Updated lens .dag file header AND the test #[ignore] attribute to use the unified two-facet framing; dissolution trigger now states 'when both facets close' so the un-ignore-after-Class-5-Gap-3 expectation isn't optimistic.

2. Added stringly-typed defense note on validate_provenance lens body: when producer wiring lands, the validate body assumes the field-path-resolves-to-real-field defense (round-trip against the *SyntaxBinding / *OpsBinding field set at HEAD); malformed paths surface as aggregate diagnostic since the EmissionRule = String type can't enforce.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 6c7c3d85 · Trigger: schedule
• Comparison: origin/main @ 617b3de1 ... review/pr-1928-6c7c3d85 @ 6c7c3d85
• Thinking: 90s wall

---

Verdict: APPROVE — The substantive change is narrow: a substrate carrier (v3.std.emission_provenance), a lens module with a fail-closed read stub (Inhabits(Empty)), integration tests that pin the Director-locked record and OptionalSourceSpan sum shape, and bootstrap regen churn. Nothing in the diff clearly breaks INVARIANTS.md (faithful carrier + forward path for provenance, boundary comments for deferred emitter wiring, fail-closed stub rather than fabricated rows), docs/modeling-discipline.md (record vs mistaken coproduct for orthogonal rule/span is explicit; sum types that appear are documented; scaffold carries a named dissolution trigger for Class 5 Gap 3 and for producer wiring), CODING.md, or TESTING.md (integration-layer compile_to_dag for structural cementing matches the doc’s exception for pipeline/integration tests; the #[ignore]d test documents why and what unblocks it).

Exploratory (optional): In emission_provenance_lens_test.rs, the LENS_FIXTURE_SOURCE helpers concat_provenance / branch_provenance return only a / l, unlike src/v3/lenses/emission_provenance.dag which uses concat on both arms. The test is ignored today, but when Gap 3 closes it may be worth matching the real monoid bodies so the eventual compile receipt does not drift from the canonical lens file.

[briansrls]

Addressed cursor exploratory finding (sha 6c7c3d8) in 5197e4f: fixture helpers concat_provenance/branch_provenance now use concat(a, b) / concat(l, r) on both arms, matching the canonical lens .dag file. Avoids drift when Class 5 Gap 3 closes and the #[ignore]d test un-ignores.

Same commit also fixes a follow-on test failure that surfaced after the bootstrap regen at 6c7c3d8: switched the carrier-shape tests to query generated_full_bootstrap_dag() directly (the new declarations are now part of the bootstrap; re-compiling the carrier source collided with duplicate-declaration errors).

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 6c7c3d85 · Trigger: schedule
• Thinking: 280s wall

---

BLOCKING (2)

Root Cause

• src/v3/std/emission_provenance.dag Line-number semantics were introduced as a bare primitive instead of a positive line-coordinate carrier → type emitted_line as PositiveInt or a dedicated positive EmittedLine and cement that field type in the test.
• src/v3/std/emission_provenance.dag The optional-span absence semantics are described but not classified under the coproduct receipt process → add the terminal or scaffold classification with its ledger or trigger at the declaration.

⚠️ Two substrate-shape issues should be fixed before this carrier becomes the downstream pattern.

[briansrls]

Both codex BLOCKING findings (sha 6c7c3d8) addressed in cbab07e:

Finding 1 (line-number positive-int carrier): emitted_line: Int → emitted_line: PositiveInt (dsl/std/integer.dag:137, Nat where gt_zero). Line coordinates are 1-indexed and positive by construction; refinement type now enforces the invariant in the type system. Cementing test emitted_line_is_positive_int_refinement_not_bare_int verifies the field type resolves to PositiveInt against the bootstrap Dag.

Finding 2 (Practice 4 receipt on OptionalSourceSpan): added 🟢 PRIMITIVE classification per docs/modeling-discipline.md#4-coproduct-dissolution — closed 2-variant typed sum, structurally distinguished (presence vs absence), irreducible. Direct analog of OptionalDiagnostic at v3.std.dimensions and the canonical 'absence-or-presence' idiom. No dissolution trigger; the two variants are exhaustive and the shape is terminal.

Bootstrap snapshots regen'd for the PositiveInt field-type change. fmt fix in e302cd5.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 6251a1e5 · Trigger: manual
• Comparison: main @ 4d59ffa3 ... session/smart-ram-167-rule-enum @ e302cd5e
• Conversation: View conversation

---

1. Story of the diff

This PR introduces v3.std.emission_provenance as the typed carrier for attributing emitted target lines: each EmissionProvenance record carries an emitted-line coordinate, a mandatory emission rule field-path, and an optional source span (src/v3/std/emission_provenance.dag:71-75). It then sketches a Lens<List<EmissionProvenance>> shape in src/v3/lenses/emission_provenance.dag, with per-behavior reads currently returning an empty witnessed list and the monoidal composition paths using concat for sequential and branch aggregation (src/v3/lenses/emission_provenance.dag:48-67). The Rust side regenerates bootstrap snapshots so the new std carrier exists in the bootstrapped DAG, and adds integration tests that pin the carrier’s declared labels plus an ignored fixture for the currently blocked top-level lens instance (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:59-194).

2. Invariant categories

1. LAYER MODEL — Finding, BLOCKING. This diff adds a substrate/std carrier, so illegal states need to be excluded at the model boundary. The comment defines the field as one-indexed — src/v3/std/emission_provenance.dag:58: // - \emitted_line: line number in emitted target output (1-indexed)— but the carrier stores it as unconstrainedsrc/v3/std/emission_provenance.dag:72: emitted_line: Int. That admits 0and negative line numbers as validEmissionProvenanceinhabitants. Because this is substrate shape, not implementation-only Rust, the carrier should use a structural line-number type/refinement/validated coordinate rather than rawInt.
2. INVARIANTS.md + modeling-discipline.md — Finding, BLOCKING. The same line violates illegal states unrepresentable / fail-closed modeling discipline: src/v3/std/emission_provenance.dag:72: emitted_line: Int makes the invalid states representable even though the model explicitly says the coordinate is one-indexed. The record-vs-coproduct decision for rule plus source_span is otherwise the right modeling direction, but the line-coordinate field still leaves a forbidden state space open.
3. CODING.md — Compliant. The new Rust test helpers are data + free functions over explicit inputs: src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:39: fn conj_field_labels(dag: &Dag, name: &str) -> Vec<String> { and src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:49: fn disj_variant_labels(dag: &Dag, name: &str) -> Vec<String> {; no hidden state or object-style behavior is introduced.
4. TESTING.md — Finding, NON-BLOCKING. The active tests claim to cement the locked structural shape, but the helpers erase the field/variant payload types: src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:44: TypeConnective::Conj { children } => children.iter().map(|f| f.label.clone()).collect(), and src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:54: TypeConnective::Disj { variants } => variants.iter().map(|v| v.label.clone()).collect(),. As written, the tests would pass if rule stopped being EmissionRule, if source_span stopped being OptionalSourceSpan, or if SomeSourceSpan lost its value: SourceSpan payload. Tightening these to assert field order and child declaration/type shape would better match the stated “locked carrier” claim.
5. LOCKED DESIGN DECISIONS — Compliant. The diff explicitly calls out the relevant locked/ratified decisions instead of silently diverging: the Lens framing is documented at src/v3/std/emission_provenance.dag:5-8, and the string field-path decision for EmissionRule is documented at src/v3/std/emission_provenance.dag:21-28. I did not find a diff line that contradicts those stated decisions.
6. TRACKED vs UNTRACKED DEBT — Compliant. The behavioral deferral and scaffold are documented with bounds and dissolution triggers: producer wiring is scoped to future render_named_template instrumentation at src/v3/lenses/emission_provenance.dag:13-22, and the blocked top-level data emission_provenance_lens promotion names the Class 5 Gap 3 facets plus the trigger at src/v3/lenses/emission_provenance.dag:107-125. The ignored fixture also carries the same bounded trigger at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:167-180.

3. Verdict

REQUEST_CHANGES

The PR’s product-vs-coproduct reshape and tracked structural deferral are mostly disciplined, but the new substrate carrier admits invalid emitted-line coordinates despite declaring a one-indexed contract. Fixing that at the carrier boundary is cheaper now than after consumers start depending on EmissionProvenance.

[briansrls]

Re inline finding at src/v3/std/emission_provenance.dag:72 (sha 6c7c3d8): already addressed in cbab07e + e302cd5 (current HEAD). emitted_line is now PositiveInt (dsl/std/integer.dag:137, Nat where gt_zero); zero/negative line numbers are no longer representable. Cementing test emitted_line_is_positive_int_refinement_not_bare_int verifies the field's type resolves to PositiveInt against the bootstrap Dag.

— sent from smart-ram-167

[briansrls]

Re inline finding at src/v3/std/emission_provenance.dag:53 (sha 6c7c3d8): already addressed in cbab07e (current HEAD). OptionalSourceSpan now carries 🟢 PRIMITIVE classification per docs/modeling-discipline.md#4-coproduct-dissolution — closed 2-variant typed sum, structurally distinguished (presence vs absence), irreducible variants. Direct analog of OptionalDiagnostic at v3.std.dimensions (canonical 'absence-or-presence' idiom). No dissolution trigger because the two variants are exhaustive and the shape is terminal.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: e302cd5e · Trigger: schedule
• Comparison: origin/main @ dd5a2cf6 ... review/pr-1928-e302cd5e @ e302cd5e
• Thinking: 31s wall

---

Findings: None. The diff adds v3.std.emission_provenance (record carrier, PositiveInt, OptionalSourceSpan with explicit Practice 4 labels), lenses.emission_provenance (structural lens fns, stub read and validate with a named downstream trigger, scaffold comment with a concrete dissolution trigger for promoting the data instance), integration wiring, cementing tests against generated_full_bootstrap_dag() for locked field/refinement/shape, and an #[ignore] fixture test whose attribute documents the Class 5 Gap 3 trigger — aligned with P1 (carrier + forward wiring path), P2/P3 (single substrate carrier; stub returns honest empty lists rather than inventing provenance), and P5 (tracked scaffold, not an unnamed bridge). Regenerated bootstrap_generated*.rs churn matches introducing new .dag modules.

Verdict: APPROVE — Narrowly scoped structural landing with documented stubs and dissolution triggers; nothing in the diff clearly violates INVARIANTS.md, docs/modeling-discipline.md, CODING.md, or TESTING.md.

Exploratory observations (optional): src/v3/lenses/emission_provenance.dag:31 imports Monoid but the module body never references it — worth deleting if the toolchain warns on unused imports; stale references to rust_target.rs line numbers in comments may need occasional touch-up as emit code moves.

[briansrls]

openai-pro NON-BLOCKING finding (sha da335d1) addressed in 6d5d908: fixture now mirrors production lens exactly.

• Added Empty to fixture's std.list import (was using Empty at multiple sites without importing it).
• read_provenance body: Inhabits(Empty) → Inhabits(_seed_list_provenance_empty()) to match production's seeded form.

Now when Class 5 Gap 3 closes, removing the #[ignore] is sufficient — no fixture/production mismatch hidden behind the substrate-cascade trigger.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 0b0db483 · Trigger: schedule
• Thinking: 236s wall

---

BLOCKING (1)

Root Cause

• src/v3/std/emission_provenance.dag emitted_line was grounded through the conceptual integer module instead of the existing PositiveInt authority → import std.types { PositiveInt }, update the nearby dsl/std/types.dag reference/definition note, and mirror the import fix in the lens test fixture.

⚠️ One substrate authority edge needs to be corrected before this lands.

[briansrls]

BLOCKING: PositiveInt is imported from std.integer even though its canonical declaration is std.types (dsl/std/types.dag:244), so this new substrate carrier records the wrong authority edge and violates M2/M9 single-authority grounding.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 6d5d9087 · Trigger: schedule
• Comparison: origin/main @ fd1a65ab ... review/pr-1928-6d5d9087 @ 6d5d9087
• Thinking: 32s wall

---

Verdict: APPROVE — clean substrate slice.

This adds the EmissionProvenance carrier (src/v3/std/emission_provenance.dag) and the structural Lens<List<EmissionProvenance>> (src/v3/lenses/emission_provenance.dag), with the four expected co-edits all present:

• Bootstrap authority entry (src/v3/std/bootstrap_authority.dag:62).
• Parse-corpus manifest hash bumped + new line (parse_corpus_manifest.txt:34,43).
• SG-0 census expected-test entry (sg0_census_test.rs:352-358).
• Generated bootstrap snapshots regenerated.

Modeling discipline: fail-closed shape is real, not cosmetic — rule: EmissionRule is mandatory (record over coproduct, per the cited reshape from PR #1910), and OptionalSourceSpan's two-arm typed sum follows the OptionalDiagnostic precedent rather than introducing a generic Option. emitted_line: PositiveInt lifts the 1-indexed contract into the type system. read_provenance returns Inhabits(Empty) — an honest "no attribution authored yet" rather than fabricating a span.

Scaffold accounting: the deferred fixture-bound data emission_provenance_lens instance has all three properties — documented (header comment in the lens .dag and the test ignore-reason), bounded (Class 5 Gap 3), named dissolution trigger (un-ignore when sum-variant + fn-typed ValueBody carriers land). The production_lens_module_compiles_cleanly smoke test plugs the gap that the #[ignore]d cementing test would otherwise leave (drift in the production .dag would be silent).

Nothing in the diff to flag against INVARIANTS / modeling-discipline / CODING / TESTING.

[briansrls]

Re inline finding at src/v3/std/emission_provenance.dag:60 (sha da335d1): already addressed at HEAD in cbab07e. OptionalSourceSpan carries the 🟢 PRIMITIVE classification at line 46 of the current file:

// **Practice 4 classification: 🟢 PRIMITIVE** — closed 2-variant typed sum, structurally distinguished (presence vs absence), irreducible variants. Direct analog of OptionalDiagnostic at v3.std.dimensions and the canonical "absence-or-presence" idiom. No dissolution trigger — the two variants are exhaustive and the shape is terminal.

Per docs/modeling-discipline.md#4-coproduct-dissolution: 🟢 PRIMITIVE shapes don't require a dissolution trigger because they're already terminal. The classification + rationale satisfies the coproduct dissolution discipline.

— sent from smart-ram-167

[briansrls]

Re inline finding at src/v3/std/emission_provenance.dag:32 (sha da335d1): finding is incorrect. The canonical declaration of PositiveInt is dsl/std/integer.dag:137 (type PositiveInt = Nat where gt_zero), NOT dsl/std/types.dag.

dsl/std/types.dag:260 documents the migration explicitly:

// 'NonNegativeInt' and 'PositiveInt' migrated to Nat-based shapes per S9

And dsl/std/types.dag:266 gives the canonical import shape:

// 'import std.integer { NonNegativeInt, PositiveInt }'

My import import std.integer { PositiveInt } matches the canonical authority and the precedent at src/v3/std/approximate_field.dag:19 (which imports the same way). Single-authority grounding is on std.integer, not std.types — types.dag only carries the migration receipt.

— sent from smart-ram-167

[briansrls]

Re BLOCKING review on sha 0b0db48 (codex): finding is incorrect — same as the inline finding at line :32 already replied to. Re-stating for the top-level review:

PositiveInt canonical declaration is at dsl/std/integer.dag:137 (type PositiveInt = Nat where gt_zero), NOT dsl/std/types.dag. types.dag:260 carries the migration receipt for the S9 move ('NonNegativeInt' and 'PositiveInt' migrated to Nat-based shapes per S9); types.dag:266 documents the canonical import shape as import std.integer { NonNegativeInt, PositiveInt }.

grep 'type PositiveInt' dsl/std/ returns only dsl/std/integer.dag:137 — there is no PositiveInt declaration in types.dag. My import is correct; the precedent at src/v3/std/approximate_field.dag:19 matches.

If the codex reviewer's authority for 'canonical home is types.dag' is something other than the S9 migration receipt I'm citing, please surface — but on substrate-state-grep at HEAD, integer.dag is the single authority.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 6d5d9087 · Trigger: manual
• Comparison: main @ e9152732 ... session/smart-ram-167-rule-enum @ 6d5d9087
• Conversation: View conversation

---

1. Story of the diff

This PR introduces std.emission_provenance as a bootstrap-authoritative std module and regenerates the bootstrap snapshots around the new declarations. The new carrier models provenance as a product record, not a coproduct: each emitted line has a positive line number, a mandatory EmissionRule, and an optional SourceSpan via OptionalSourceSpan (src/v3/std/emission_provenance.dag:59-86). The lens side adds lenses.emission_provenance, which currently lands the structural functions for List<EmissionProvenance>: read returns an empty witness stub, sequential/branch compose by concat, iteration is identity, and validation is a placeholder while producer wiring is deferred (src/v3/lenses/emission_provenance.dag:43-104). The tests wire a new integration module, assert the bootstrap carrier and optional payload shapes, smoke-compile the production lens file, and record the new hand-authored test in the SG-0 census (src/v3/compiler/tests/integration.rs:60-61, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:39-187, src/v3/compiler/tests/integration/sg0_census_test.rs:352-358).

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this is substrate-touching, and the new carrier is modeled on the Dag as declared std types rather than Rust-only implementation state: EmissionProvenance is a record with typed fields at src/v3/std/emission_provenance.dag:82-86, and the module is added to bootstrap authority at src/v3/std/bootstrap_authority.dag:62.

1. INVARIANTS.md + modeling-discipline.md.

Finding (NON-BLOCKING) — P1 Documentation describes live state / Modeling Faithfulness. src/v3/std/emission_provenance.dag:7-8 says: “This file authors the typed carrier; the per-Behavior Lens<List<EmissionProvenance>> instance lives at lenses.emission_provenance.” But the landed lens file says the actual data emission_provenance_lens: Lens<List<EmissionProvenance>> declaration is not top-level production state: “the data emission_provenance_lens... instance declaration is authored as a TEST FIXTURE” and “rather than as a top-level data item here” at src/v3/lenses/emission_provenance.dag:106-113. The current live state is structural lens functions plus a fixture-bound/ignored instance, not a production Lens<C> data instance; update the std/lens headers to say the instance is fixture-bound until Class 5 Gap 3 closes so consumers do not treat a non-existent production declaration as an authority.

1. CODING.md.

Compliant — the production lens logic follows data + free functions rather than object/method state: read_provenance, concat_provenance, branch_provenance, iterate_provenance, and validate_provenance are plain declared functions at src/v3/lenses/emission_provenance.dag:56-104; the Rust additions are test wiring and assertions, not new production object surfaces.

1. TESTING.md.

Compliant — the PR adds active structural tests at the right seam: carrier field labels and field declaration IDs are asserted at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:61-99, the production lens file is smoke-compiled at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:110-124, and OptionalSourceSpan.SomeSourceSpan.value is checked against canonical SourceSpan at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:161-176. The ignored fixture-bound lens binding test is explicitly bounded with a Class 5 Gap 3 dissolution trigger at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:255-268.

1. LOCKED DESIGN DECISIONS.

Compliant — the diff follows the referenced locked direction by choosing per-Behavior Lens<C> analysis rather than runtime instrumentation (src/v3/std/emission_provenance.dag:5-8, src/v3/lenses/emission_provenance.dag:7-11) and by keeping EmissionRule as the field-path reason carrier rather than adding a separate nominal rule enumeration (src/v3/std/emission_provenance.dag:21-42).

1. TRACKED vs UNTRACKED DEBT.

Compliant — the new temporary shapes are documented, bounded, and carry triggers: the empty read_provenance stub names producer wiring as the lift condition at src/v3/lenses/emission_provenance.dag:47-55, and the top-level lens-instance scaffold records the Class 5 Gap 3 facets plus the promotion trigger at src/v3/lenses/emission_provenance.dag:106-134. The hand-authored test debt is also recorded in the SG-0 census at src/v3/compiler/tests/integration/sg0_census_test.rs:352-358.

3. Verdict

APPROVE_WITH_COMMENTS

No blocking substrate-shape issue: the carrier is modeled as the intended record, the optional span absence is typed, and the deferred lens work is tracked. The one cleanup I would make before or shortly after merge is to align the headers with live state: this PR lands structural lens functions and a fixture-bound instance scaffold, not a production Lens<List<EmissionProvenance>> instance.

[briansrls]

openai-pro NON-BLOCKING P1 finding (sha 6d5d908) addressed in 094be15: aligned both file headers with live state.

• src/v3/std/emission_provenance.dag: changed "the per-Behavior Lens<List> instance lives at lenses.emission_provenance" → distinguishes between structural fns (live there) and the data instance (currently fixture-bound in the integration test, blocked on Class 5 Gap 3).
• src/v3/lenses/emission_provenance.dag: status "STRUCTURALLY TERMINAL; BEHAVIORALLY DEFERRED" → "STRUCTURALLY TERMINAL (fns); LENS-INSTANCE FIXTURE-BOUND; BEHAVIORALLY DEFERRED" with a "What lives here vs. the test fixture" breakdown.

Bootstrap snapshots regen'd (header edits shifted byte offsets); parse manifest refreshed.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 094be156 · Trigger: manual
• Comparison: main @ fd1a65ab ... session/smart-ram-167-rule-enum @ 094be156
• Conversation: View conversation

---

1. Story of the diff

This PR adds a substrate carrier for emission provenance and a structural lens surface around it. The carrier lands in src/v3/std/emission_provenance.dag: EmissionRule is the field-path string carrier, OptionalSourceSpan is a typed absence/presence wrapper, and EmissionProvenance is a record with emitted_line: PositiveInt, mandatory rule, and optional source_span (src/v3/std/emission_provenance.dag:50, src/v3/std/emission_provenance.dag:67-69, src/v3/std/emission_provenance.dag:90-94). The lens file then declares the structural functions for Lens<List<EmissionProvenance>>: empty-seed/read, sequential concat, branch concat, loop identity, and aggregate validate (src/v3/lenses/emission_provenance.dag:57, src/v3/lenses/emission_provenance.dag:70-71, src/v3/lenses/emission_provenance.dag:75-79, src/v3/lenses/emission_provenance.dag:85-100, src/v3/lenses/emission_provenance.dag:117-118).

The production data emission_provenance_lens: Lens<List<EmissionProvenance>> instance is intentionally not landed yet: the diff documents it as fixture-bound because Class 5 Gap 3 blocks top-level ValueBody lowering for the needed list/sum/function-typed shape (src/v3/lenses/emission_provenance.dag:120-147). The rest of the diff wires the new std file into bootstrap authority and parse-corpus coverage (src/v3/std/bootstrap_authority.dag:62, src/v3/compiler/tests/integration/parse_corpus_manifest.txt:43), regenerates both bootstrap snapshots (src/v3/compiler/src/bootstrap_generated.rs:26209-26251, src/v3/compiler/src/bootstrap_generated_without_parse_surface.rs:26209-26251), and adds integration tests for the carrier shape, production lens parsing, and the fixture-bound future lens instance (src/v3/compiler/tests/integration.rs:60-61, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:39-282).

2. Invariant categories

1. LAYER MODEL — Compliant. This is a substrate-touching PR, and the carrier shape uses product modeling for orthogonal facts rather than reintroducing the earlier coproduct compression: EmissionProvenance is a record with emitted_line, rule, and source_span fields (src/v3/std/emission_provenance.dag:87-94), while absence of source span is modeled as the typed NoSourceSpan | SomeSourceSpan { value: SourceSpan } carrier (src/v3/std/emission_provenance.dag:67-69). chatgpt-review-0f4e94a3-54ee-4a…
2. INVARIANTS.md + modeling-discipline.md — Compliant. The diff handles the main modeling-pressure points explicitly: emitted_line: PositiveInt makes non-positive emitted-line coordinates unrepresentable (src/v3/std/emission_provenance.dag:73-77, src/v3/std/emission_provenance.dag:91), mandatory rule makes “rule absent” structurally impossible (src/v3/std/emission_provenance.dag:78-81, src/v3/std/emission_provenance.dag:92), and the record/product form dissolves the prior origin coproduct because rule and span can both apply to the same emitted line (src/v3/std/emission_provenance.dag:18-27, src/v3/std/emission_provenance.dag:87-89). chatgpt-review-0f4e94a3-54ee-4a…

chatgpt-review-0bd87144-fc54-49…

1. CODING.md — Compliant. The new implementation-facing shape follows data + free functions rather than method/object scaffolding: the lens module exposes standalone functions such as read_provenance, concat_provenance, branch_provenance, iterate_provenance, and validate_provenance (src/v3/lenses/emission_provenance.dag:70, src/v3/lenses/emission_provenance.dag:75, src/v3/lenses/emission_provenance.dag:85, src/v3/lenses/emission_provenance.dag:96, src/v3/lenses/emission_provenance.dag:117). chatgpt-review-2b19e72f-5169-46…
2. TESTING.md — Finding, NON-BLOCKING. The carrier tests say they are locking exact record/sum shapes, but the assertions collapse labels through HashSet, which masks duplicate fields/variants and extra copies. The field-shape test builds a set instead of asserting arity: src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:62: let expected_labels: HashSet<&str> = ["emitted_line", "rule", "source_span"] and src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:65: let actual_labels: HashSet<&str> = labels.iter().copied().collect();. The same pattern appears for OptionalSourceSpan at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:148-149. Testing principle: behavior-driven/cost-of-change tests should pin the actual published contract; here the published contract is “three fields” and “two variants,” so the tests should also assert children.len() == 3 and variants.len() == 2 before or alongside the set comparison. chatgpt-review-9bcb03db-dee6-43…
3. LOCKED DESIGN DECISIONS — Compliant. The diff references the Director-ratified choices and implements them without silent divergence: per-Behavior lens framing is called out in the lens header (src/v3/lenses/emission_provenance.dag:21-25), and the top-level lens instance’s non-production status is explicit rather than disguised as complete (src/v3/std/emission_provenance.dag:10-16, src/v3/lenses/emission_provenance.dag:120-147).
4. TRACKED vs UNTRACKED DEBT — Compliant. The scaffolds I found are tracked: behavioral producer wiring is bounded as a downstream slice (src/v3/lenses/emission_provenance.dag:27-36), the current read_provenance empty output is labeled a stub with a producer-wiring trigger (src/v3/lenses/emission_provenance.dag:61-69), the fixture-bound lens instance has a named Class 5 Gap 3 dissolution trigger (src/v3/lenses/emission_provenance.dag:129-147), and the ignored fixture test carries the same trigger in its attribute (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:255-268).

3. Verdict

APPROVE_WITH_COMMENTS

The substrate/lens modeling choices look sound: the diff avoids the old coproduct compression, makes the mandatory provenance rule structural, and documents the fixture-bound lens instance as tracked debt rather than pretending it is production-complete. The only issue I would ask to follow up is the non-blocking test-strength gap: the HashSet assertions should also pin field/variant arity so the “locked shape” tests actually fail on duplicated or extra structural members.

[briansrls]

openai-pro NON-BLOCKING TESTING finding (sha 094be15) addressed in c09e3a2: added exact-arity assertions before the HashSet membership comparison in both carrier-shape tests:

• EmissionProvenance: children.len() == 3
• OptionalSourceSpan: variants.len() == 2

A record/sum with duplicated structural members would pass the set comparison; the arity check catches it. Failure messages include the full label list for visibility.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 094be156 · Trigger: schedule
• Comparison: origin/main @ fd1a65ab ... review/pr-1928-094be156 @ 094be156
• Thinking: 47s wall

---

Reviewed the non-bootstrap_generated* diff: new std carrier (emission_provenance.dag), bootstrap authority entry, lens module (lenses/emission_provenance.dag), integration tests and wiring, SG-0 census line, parse corpus manifest row, plus mechanical regen churn in the two bootstrap files.

Findings: None. The substrate carrier matches the locked record-vs-coproduct rationale (P1 / coproduct-vs-coordinate), OptionalSourceSpan is explicitly classified in-DAG, fixture-bound lens instance and stub read/validate bodies are documented with named dissolution triggers (Class 5 Gap 3 for top-level data, separate slice for emitter wiring), which satisfies the tracked-scaffold bar in your review rules. New hand-authored test Rust is registered in the SG-0 census. Nothing here violates fail-closed as “fabricate provenance entries” (empty list + NoDiagnostic is an honest stub, not a plausible fake row). CODING.md / TESTING.md stance for new code: integration/compile smoke for a standalone lens file matches the cited precedent pattern; test names state concrete behaviors.

Verdict: APPROVE — The change is narrowly scoped to the EmissionProvenance carrier and structural Lens<List<…>> shape with explicit deferrals and receipts; no concrete rubric violations surfaced in the reviewed diff.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: c09e3a21 · Trigger: manual
• Comparison: main @ fd1a65ab ... session/smart-ram-167-rule-enum @ c09e3a21
• Conversation: View conversation

---

1. Story of the diff

This PR lands emission provenance as a declared substrate carrier rather than as emitter-side instrumentation or a parallel per-line table. The new src/v3/std/emission_provenance.dag introduces EmissionRule, OptionalSourceSpan, and the product-shaped EmissionProvenance record, with emitted_line: PositiveInt, mandatory rule, and optional source attribution. src/v3/lenses/emission_provenance.dag then gives the structural Lens<List<EmissionProvenance>> functions: per-behavior read is currently an empty-list stub, sequential and branch composition use concat, iteration is identity, and aggregate validation is deferred.

The bootstrap authority, parse manifest, SG-0 census, and generated bootstrap snapshots are updated so the new std carrier is part of the bootstrapped Dag. The Rust integration test module actively pins the carrier shape and compiles the production lens module, while the full fixture-bound lens instance is present but ignored with an explicit Class 5 Gap 3 dissolution trigger.

2. Invariant categories

1. LAYER MODEL — Compliant. This is substrate-touching, and the core carrier is modeled as a record rather than a compressed origin coproduct: src/v3/std/emission_provenance.dag:90-94 declares EmissionProvenance { emitted_line, rule, source_span }, and src/v3/std/bootstrap_authority.dag:62 adds the file to bootstrap authority.
2. INVARIANTS.md + modeling-discipline.md — Compliant. Coproduct dissolution and illegal-state discipline are handled in the carrier: src/v3/std/emission_provenance.dag:87-94 states the rule/span axes are orthogonal and encodes them as a product, while src/v3/std/emission_provenance.dag:67-69 uses a typed OptionalSourceSpan instead of fabricating a span.
3. CODING.md — Compliant. The added Rust is test-harness code, and it stays in the data + free-functions style: tests call explicit inputs such as generated_full_bootstrap_dag() at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:51 and cached_compile_to_dag(...) at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:130-133, without adding production-side methods, hidden state, or object-style APIs.
4. TESTING.md — Finding (NON-BLOCKING). The production carrier declares the locked optional shape as NoSourceSpan | SomeSourceSpan { value: SourceSpan } at src/v3/std/emission_provenance.dag:67-69, but the test only searches the SomeSourceSpan payload for a field named value at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:184-188. It does not assert that SomeSourceSpan has exactly one payload field, and it does not inspect the NoSourceSpan payload at all. A future drift to SomeSourceSpan { value, extra } or NoSourceSpan { value: SourceSpan } would still pass this test even though the test name promises the locked payload shape.
5. LOCKED DESIGN DECISIONS — Compliant. The PR explicitly names the locked/director decisions it depends on: per-behavior Lens<C> provenance rather than runtime instrumentation at src/v3/lenses/emission_provenance.dag:21-25, and EmissionRule = String as field-path identity at src/v3/std/emission_provenance.dag:29-36 plus the declaration at src/v3/std/emission_provenance.dag:50. I do not see an implicit divergence from a locked decision in the diff.
6. TRACKED vs UNTRACKED DEBT — Compliant. The temporary shapes are documented, bounded, and have dissolution triggers. The read stub is marked as returning empty provenance until producer wiring lands at src/v3/lenses/emission_provenance.dag:61-69, and the fixture-bound lens instance scaffold names the Class 5 Gap 3 blocker plus the promotion trigger at src/v3/lenses/emission_provenance.dag:120-148. The ignored fixture test carries the same trigger at src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:273-286.

3. Verdict

APPROVE_WITH_COMMENTS. The substrate carrier shape, bootstrap integration, and scaffold accounting are sound in this diff. The only issue I found is a non-blocking test-ratchet gap around exact OptionalSourceSpan payload arity; the current production carrier is correct, but the regression test should pin both variant payload shapes exactly.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: c09e3a21 · Trigger: schedule
• Comparison: origin/main @ fd1a65ab ... review/pr-1928-c09e3a21 @ c09e3a21
• Thinking: 34s wall

---

Findings

None. The substantive diff adds a grounded substrate carrier (EmissionProvenance as a product type with a documented reshape from the old coproduct), registers it in bootstrap_authority.dag and the parse corpus, and implements a Lens<List<EmissionProvenance>> shape in src/v3/lenses/emission_provenance.dag with explicit 🟡 stub/scaffold notes and named dissolution triggers (Class 5 Gap 3) in both the lens file and emission_provenance_lens_test.rs (e.g. the #[ignore] reason and LENS_FIXTURE_SOURCE comments). OptionalSourceSpan includes a 🟢 classification with rationale per docs/modeling-discipline.md Practice 4. The new integration test is registered in sg0_census_test.rs with a PR-scoped receipt. The large bootstrap_generated*.rs churn is the expected mechanical follow-on from a new std module, not a separate design issue in this diff.

Verdict

APPROVE — The change matches the rubric: single-authority carrier in std, lens as declared analysis over that carrier, fail-closed framing for the empty read stub (no fabricated provenance rows), and scaffolds are bounded, documented, and tied to a concrete trigger. No invariant or doc violation stood out in the diff.

Exploratory (optional): After producer wiring exists, you may want a follow-up that tightens validate_provenance beyond NoDiagnostic; that is already explicitly deferred in the diff and is not a blocker for this structural slice.

[briansrls]

openai-pro NON-BLOCKING TESTING finding (sha c09e3a2) addressed in a4a183b (+ 7a122e2 fmt):

• SomeSourceSpan payload arity: some_children.len() == 1 before find()-ing the value field — SomeSourceSpan { value, extra } would now fail.
• NoSourceSpan payload: when connective is Conj, assert children.len() == 0 — payload-bearing NoSourceSpan { value: SourceSpan } would collapse the absence/presence distinction with SomeSourceSpan and is now caught explicitly.

Other connective kinds (unit/Top synthesized for unit variants) remain acceptable — only payload-bearing NoSourceSpan fails.

— sent from smart-ram-167

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 7a122e2e · Trigger: manual
• Comparison: main @ fd1a65ab ... session/smart-ram-167-rule-enum @ 7a122e2e
• Conversation: View conversation

---

1. Story of the diff

This PR introduces emission provenance as a typed substrate carrier instead of leaving “what emitted this line?” as an emitter-side convention. The new v3.std.emission_provenance module defines EmissionRule as the stable field-path reason, models source-span presence explicitly with OptionalSourceSpan, and makes EmissionProvenance a product carrier with emitted_line: PositiveInt, mandatory rule, and optional source_span (src/v3/std/emission_provenance.dag:43-50, src/v3/std/emission_provenance.dag:67-69, src/v3/std/emission_provenance.dag:90-94). It is then admitted to bootstrap authority (src/v3/std/bootstrap_authority.dag:62) and regenerated into the bootstrap snapshots, including the new named declarations and payload carriers (src/v3/compiler/src/bootstrap_generated.rs:26209-26277, src/v3/compiler/src/bootstrap_generated.rs:62745-62774).

The lens side lands the structural functions for Lens<List<EmissionProvenance>>: an explicit typed Empty seed, per-behavior read currently returning empty provenance, list concatenation for sequential and branch composition, loop identity for static emission, and a no-op aggregate validator while producer wiring is deferred (src/v3/lenses/emission_provenance.dag:57-118). The PR is careful not to pretend the full production lens instance is landed: both the std carrier and lens file document that the top-level data emission_provenance_lens instance is fixture-bound until Class 5 Gap 3 closes (src/v3/std/emission_provenance.dag:10-16, src/v3/lenses/emission_provenance.dag:120-148). Tests then pin the carrier’s field arity/types, the optional span variant payload shape, and active compilation of the production lens .dag file (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:39-134, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:136-239).

2. Invariant categories

1. LAYER MODEL — Compliant. This is substrate-touching: the new carrier lives in src/v3/std and is bootstrapped. The shape honors substrate modeling discipline by making the attribution record a product, not a lossy origin coproduct, and by making “rule absent” structurally impossible through the mandatory rule field (src/v3/std/emission_provenance.dag:18-27, src/v3/std/emission_provenance.dag:90-94). Bootstrap authority and generated declarations are updated in the same diff, so the new std file is not an orphan declaration (src/v3/std/bootstrap_authority.dag:62, src/v3/compiler/src/bootstrap_generated.rs:26251-26266).
2. INVARIANTS.md + modeling-discipline.md — Compliant. Coproduct dissolution / modeling faithfulness is handled by the record form: the diff explicitly states that rule and span are orthogonal facts and models them as coordinates (src/v3/std/emission_provenance.dag:87-94). Fail-closed is also respected at the span boundary: NoSourceSpan is a real variant rather than a fabricated source span, while SomeSourceSpan carries the canonical SourceSpan payload (src/v3/std/emission_provenance.dag:60-69). Rubric reference applied from the supplied invariants/modeling docs. chatgpt-review-c8197a1c-8c5a-48…

chatgpt-review-1c1735c6-10a5-4e…

1. CODING.md — Compliant. The production lens surface uses data + top-level functions rather than attaching behavior to a god object: _seed_list_provenance_empty, read_provenance, concat_provenance, branch_provenance, iterate_provenance, and validate_provenance are declared as independent functions over their explicit inputs (src/v3/lenses/emission_provenance.dag:57-118). The Rust additions are tests only, and they read the DAG through explicit query/accessor surfaces instead of introducing new production Rust state (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:51-60, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:147-156). chatgpt-review-ff8b4c60-cacd-49…
2. TESTING.md — Compliant. The diff adds active integration coverage and wires it into the test module list (src/v3/compiler/tests/integration.rs:60-61). The tests are behavior/contract-shaped: one pins EmissionProvenance as exactly the locked three-field carrier and checks field types, one pins OptionalSourceSpan as exactly the two-arm shape with SomeSourceSpan.value: SourceSpan, and one actively compiles the production lens file so syntax/type drift is caught (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:39-134, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:136-239). The ignored fixture test is not silent coverage theater: it names the blocker and dissolution trigger in the attribute itself (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:307-320). chatgpt-review-20b7b37b-5d34-43…
3. LOCKED DESIGN DECISIONS — Compliant. The diff references Director-ratified/locked decisions and follows them rather than diverging: Q1 chooses per-Behavior Lens<C> analysis instead of runtime instrumentation (src/v3/lenses/emission_provenance.dag:21-36), and Q3 keeps the stable emission reason as a field-path string rather than a nominal enum (src/v3/std/emission_provenance.dag:29-36, src/v3/std/emission_provenance.dag:43-50). The tests also pin the locked carrier shape and field types rather than just checking name presence (src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:40-50, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:65-107).
4. TRACKED vs UNTRACKED DEBT — Compliant. The two temporary shapes I see are tracked bridges, not unbounded debt. The read/validate bodies are explicitly marked as a behavior-deferred stub until emitter producer wiring records real field-path provenance (src/v3/lenses/emission_provenance.dag:27-36, src/v3/lenses/emission_provenance.dag:61-69, src/v3/lenses/emission_provenance.dag:110-118). The missing top-level data emission_provenance_lens instance is documented with the exact blocking substrate gap, both known facets, bounds, and a named dissolution trigger: when Class 5 Gap 3 closes for top-level list/sum-variant carriers and fn-typed data-body fields, the fixture promotes without lens-shape changes (src/v3/lenses/emission_provenance.dag:120-148, src/v3/compiler/tests/integration/emission_provenance_lens_test.rs:307-320). The SG-0 hand-authored test entry is also explicitly accounted for (src/v3/compiler/tests/integration/sg0_census_test.rs:352-358).

3. Verdict

APPROVE

I did not find a blocking or non-blocking finding that is both tied to the diff and not already tracked by the PR’s own scaffold receipts. The substrate carrier shape is faithful and typed, the bootstrap/manifest/test wiring is present, and the remaining behavioral work is bounded with named dissolution triggers rather than smuggled in as steady state.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: 7a122e2e · Trigger: schedule
• Comparison: origin/main @ fd1a65ab ... review/pr-1928-7a122e2e @ 7a122e2e
• Thinking: 37s wall

---

APPROVE

Diff is small in substance despite the bootstrap regen volume, and I don’t see a concrete violation of the pinned modeling, coding, or testing docs. The new substrate carrier in src/v3/std/emission_provenance.dag uses a single-authority record shape for the two orthogonal facts, the OptionalSourceSpan coproduct is explicitly classified and terminal, the lens shape in src/v3/lenses/emission_provenance.dag is documented as tracked scaffold with a bounded dissolution trigger, and the same-PR structural consumers/tests are present.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 7a122e2e · Trigger: schedule
• Thinking: 308s wall

---

Non-blocking — Strengths

• src/v3/lenses/emission_provenance.dag The Class 5 Gap 3 deferral is documented, bounded to the top-level data-body boundary, and names the dissolution trigger.

⚠️ Prior blocking comment 3198478491 remains unresolved; no additional new blocking concerns found.