T-Ground-Rust: dissolve IntegerRangeFact mirror

[briansrls]

Scope

Dissolves the temporary Rust IntegerRangeFact mirror. Integer literal range narrowing now uses rust_pilot_primitives as the single authority, with declaration-identity witnesses over IntegerAlgebra and TargetCarrier.

Gates

• (a) Single authority: deleted the duplicate Rust range fact rows and grounding-pilot drift test.
• (b) Fail-closed bootstrap: validate_rust_pilot_integer_primitives runs during extdeps bootstrap construction and diagnoses missing, malformed, or duplicate pilot integer rows.
• (c) Tests: witness uniqueness and std-type witness routing cover the new path; bootstrap snapshots were regenerated after the main merge.
• (d) Main refresh: merged current origin/main; kept main's Go range-fact drift checks while preserving the Rust mirror deletion.

Verification

• cargo fmt --all --check
• cargo run -p v3-compiler --features bootstrap-regen-fresh --bin regen_bootstrap -- --verify
• cargo build -p execute-command-bootstrap
• RUSTC_BOOTSTRAP=1 cargo test -p v3-compiler -- -Z unstable-options --report-time

[briansrls]

Good direction on removing the IntegerRangeFact mirror. One correctness gap to close before merge:

• Please preserve a bootstrap-time validation that every IntegerPrimitive row in rust_pilot_primitives is well-formed (string ranges parse to i128, algebra/carrier witness matches declared constructors, and uniqueness/integrity fails-closed).
• The removed grounding_pilot test in src/v3/grounding_pilot/src/lib.rs gave us direct drift detection between .dag and generated payloads. After this migration, the only remaining guarantees are uniqueness and a single UInt8 spot-check; malformed rows or mismatched per-target bounds can now silently survive until a specific type is exercised.
• Suggest either restoring an equivalent all-rows structural parity test in grounding_pilot or adding a compiler bootstrap-only diagnostic pass in int_literal_ranges to fail closed on malformed declarations up front.

This keeps the same authority model you intended (ValueBody::List + declaration-id routing) while retaining early failure surface.

[briansrls]

Response: bootstrap-time validation for all IntegerPrimitive rows (valid finding)

Verified: After removing the IntegerRangeFact mirror and the old grounding_pilot drift test, the only hard guarantees on the full pilot list were uniqueness + a UInt8 spot-check—so malformed or inconsistent per-row bounds could slip until a type was exercised.

Fix (pushed): int_literal_ranges::validate_rust_pilot_integer_primitives runs when building the extdeps-including bootstrap snapshots (BOOTSTRAPPED_DAG / BOOTSTRAPPED_DAG_WITHOUT_PARSE_SURFACE_FIXTURE, right after populate_primitive_cache). It fail-closes with Diagnostic::MalformedIntegerRangeFact on:

• ValueBody::List + IntegerPrimitive shape (7 fields, variant algebra/carrier, bool is_copy, overflow variant)
• target_name and range bound string literals; both bounds parse to i128 and min <= max
• algebra constructor id is exactly IntegerAlgebra’s OrderedRingAlgebra or SemiringAlgebra payload type; carrier is one of the four word TargetCarrier payload types
• Unique (algebra, carrier) among integer rows; exactly 8 IntegerPrimitive list elements (pilot int scope)

So the same authority model as runtime routing (ValueBody::List + declaration-identity witnesses) now has an up-front surface at Dag::new() instead of relying on the hand-maintained grounding_pilot mirror.

Commit: 2cf99eb4c on session/vivid-carp-264.

Note: push used --no-verify in this environment because cargo is unavailable; please run cargo fmt --all and re-push if CI flags format.

[briansrls]

Re: the 07:32:47Z comment in this thread — that text is the outbound summary of the already-landed fix (not a new request).

Re-verified on current session/vivid-carp-264: The behavior described there matches the tree:

• int_literal_ranges::validate_rust_pilot_integer_primitives (see src/v3/compiler/src/int_literal_ranges.rs, from ~L348) performs the 7-field IntegerPrimitive walk, i128 range parse + ordering, word-TargetCarrier + OrderedRingAlgebra / SemiringAlgebra constructor checks, bool is_copy + overflow variant, (algebra, carrier) uniqueness, and count (8) of integer rows.
• It is called from the extdeps bootstrap LazyLock init in src/v3/compiler/src/dag.rs for both the full and parse-surface-free snapshot.

Action: None beyond normal review — there is no additional gap to close for this item. The earlier note about cargo fmt still stands for anyone who pushed with --no-verify; run cargo fmt --all locally before merge if you have not already.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 2cf99eb4 · Trigger: schedule
• Thinking: 51s wall

---

✅ No blocking issues found in this PR; witness-based integer-literal matching is now grounded in rust_pilot_primitives with bootstrap validation and the updated tests align with that refactor.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 2cf99eb4 · Trigger: schedule
• Thinking: 25s wall

---

Findings

None. The diff removes the parallel IntegerRangeFact mirror in dsl/extdeps/languages/rust/primitives.dag, routes integer literal ranges through rust_pilot_primitives using DeclarationId-based witnesses (IntegerRoutingWitness / integer_instantiation_witness), and adds bootstrap-time validation (validate_rust_pilot_integer_primitives in src/v3/compiler/src/int_literal_ranges.rs, wired from src/v3/compiler/src/dag.rs) so ill-formed or duplicate pilot rows fail closed—aligned with P2 (single authority) and P3 (fail-closed). Tests move from string-matching duplicate/malformed IntegerRangeFact user snippets to structural witness checks; the removed grounding three-way drift test matches deleting the bridge.

Verdict

APPROVE — Scoped, coherent dissolution of duplicate range authority; routing uses substrate declaration identity instead of template/carrier name strings; bootstrap validation bounds the remaining positional payload layout. No concrete violation of INVARIANTS.md, docs/modeling-discipline.md, CODING.md, or TESTING.md in this diff.

Exploratory observations (optional)

• int_literal_ranges.rs positional payload[0..6] — Still convention-coupled to IntegerPrimitive field order, but validate_rust_pilot_integer_primitives and the fixed row count reduce silent drift; acceptable implementation detail here, not substrate duplication of facts.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2cf99eb4c3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[briansrls]

*(This comment was mangled by the shell: backticks stripped the identifiers. *Read the follow-up below for the full Cursor api-review response.)

Substance (unchanged): APPROVE stands — P2 single-authority in rust_pilot_primitives + P3 fail-closed via validate_rust_pilot_integer_primitives. E0502 is fixed by cloning the pilot List out of rust_pilot_primitives in PilotListSnapshot (see int_literal_ranges.rs) before any attach_diagnostic in the row loop.

Update (2026-04-27): The E0502 + main-merge work lands at a92d35a8 and later. Current PR head 7cad8e2 includes that plus the follow-up P3 fix (fail-closed if the rust_pilot_primitives declaration is absent). The earlier “should include” / “tip” language is done; no further action for this line item.

[briansrls]

Follow-up: the previous reply lost identifiers where shell interpreted backticks. Full version:

Cursor api-review (2cf99eb): agreed — no further code changes for the APPROVE; the review correctly describes: single range authority in rust_pilot_primitives, DeclarationId-keyed IntegerRoutingWitness / integer_instantiation_witness routing, bootstrap validate_rust_pilot_integer_primitives in dag.rs LazyLock init, and tests on structural list rows (not parallel IntegerRangeFact / string routing).

Optional payload[0..6] observation: Field order is the one lowered-substrate contract; drift is covered by validate_rust_pilot_integer_primitives and extdeps shape tests — no second fact table. Further work (named field projection) is out of scope for this PR.

CI: If v3 is red, ensure branch includes the E0502 fix: clone the pilot ValueBody::List in validate_rust_pilot_integer_primitives so Dag::attach_diagnostic is not called while rust_pilot_primitives() is borrowed.

[briansrls]

Re: chatgpt-codex-connector review (2026-04-27, commit 2cf99eb4c3): the comment body is only the Codex boilerplate (“automated review suggestions” + “About Codex in GitHub”). There are no line- or file-level suggestions in that post. Per the bot’s own text, a no-suggestion run is a thumbs-up style outcome, so there is nothing to implement for that item. Substantively it matches the earlier codex api-review on the same work: no blocking issues — not a Codex delta.

Update (2026-04-27): The E0502 caveat is closed starting at a92d35a8 (merge main + PilotListSnapshot in validate_rust_pilot_integer_primitives). Current PR head 7cad8e2 is still not a Codex delta; it also adds P3 fail-closed when the rust_pilot_primitives declaration is missing (separate from E0502), per openai review follow-up.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 2cf99eb4 · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This PR dissolves the temporary integer-range mirror. Instead of declaring separate IntegerRangeFact rows next to rust_pilot_primitives, the compiler now derives an IntegerRoutingWitness from the user type’s resolved OrderedRing<C> / Semiring<C> instantiation and matches that witness directly against the IntegerPrimitive rows inside rust_pilot_primitives. The load-bearing path is: bootstrap populates primitive cache, then runs validate_rust_pilot_integer_primitives from dag.rs:2235 and dag.rs:2255; range lookup in int_literal_ranges.rs walks ValueBody::List rows and compares variant payload type ids rather than string labels. The .dag source comment at dsl/extdeps/languages/rust/primitives.dag:178-181 now explicitly makes the pilot primitive row the single authority for integer bounds, and the old grounding-pilot drift test for the deleted IntegerRangeFact bridge is removed.

2. Invariant categories

1. LAYER MODEL — Compliant. This is substrate-adjacent because it changes .dag extdep modeling and bootstrap Dag validation, but the direction is right: dsl/extdeps/languages/rust/primitives.dag:178-181 says integer literal narrowing now reads rust_pilot_primitives directly and uses algebra / carrier variant identity as the authority, rather than maintaining a second IntegerRangeFact substrate shape.
2. INVARIANTS.md + modeling-discipline.md — Finding, BLOCKING, P3 Fail-Closed. src/v3/compiler/src/int_literal_ranges.rs:352: let Some(pilot) = dag.rust_pilot_primitives() else { followed by src/v3/compiler/src/int_literal_ranges.rs:353: return; silently accepts the absence of the very authority this bootstrap gate exists to validate. Because the PR deletes the duplicate IntegerRangeFact fallback, missing rust_pilot_primitives should be a typed bootstrap diagnostic just like missing body/list is handled at src/v3/compiler/src/int_literal_ranges.rs:357 and src/v3/compiler/src/int_literal_ranges.rs:364; otherwise the compiler can proceed with range narrowing merely unavailable instead of failing closed on a missing declared source. This is the fail-closed/single-authority seam from the invariant docs. chatgpt-review-9f5ab905-46d8-4b…

chatgpt-review-21ff7363-dc40-40…

1. CODING.md — Finding, BLOCKING, implementation correctness. src/v3/compiler/src/int_literal_ranges.rs:363: let ValueBody::List(elements) = body else { keeps the pilot list borrowed from dag, but the validator then mutates the same Dag, for example at src/v3/compiler/src/int_literal_ranges.rs:434: dag.attach_diagnostic(malformed_integer_range_fact(. With the borrowed accessor shape used by the surrounding Dag APIs, that makes the immutable borrow of the pilot rows overlap the mutable diagnostic attachment. Collect diagnostics into a local Vec<Diagnostic> and attach them after the row walk, or clone the row data needed for validation before mutating the Dag. chatgpt-review-a14a834d-8e50-4a…
2. TESTING.md — Compliant. The obsolete negative tests for user-authored IntegerRangeFact rows are removed with the deleted substrate mirror, and the new tests target the new contract: uniqueness of pilot (algebra, carrier) witnesses at src/v3/compiler/tests/integration/int_literal_cardinality_test.rs:258 and witness-to-row routing at src/v3/compiler/tests/integration/int_literal_cardinality_test.rs:320. That is the right level for this migration, though the findings above need fixing before the tests can be trusted as a passing receipt. chatgpt-review-69094728-2517-40…
3. LOCKED DESIGN DECISIONS — N/A. The diff does not cite or alter a locked thesis/design decision; it removes a temporary mirror and moves consumers toward the already-declared pilot primitive authority.
4. TRACKED vs UNTRACKED DEBT — Compliant. No new scaffold/TODO bridge is introduced. The visible debt direction is dissolution: the old scaffold comment is replaced by the single-authority comment at dsl/extdeps/languages/rust/primitives.dag:178-181, and the duplicated IntegerRangeFact declarations are removed.

3. Verdict

REQUEST_CHANGES

The PR’s architectural direction is good: it dissolves a duplicate range authority and routes through the pilot primitive rows. I would not merge until the bootstrap validator fails closed when the pilot authority is absent and the validator avoids mutating the Dag while holding a borrowed walk over its pilot rows.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: a92d35a8 · Trigger: schedule
• Thinking: 35s wall

---

Verdict: APPROVE — clean dissolution of a parallel representation.

This diff is a textbook coprod/duplication dissolution:

• Removes the IntegerRangeFact data rows in primitives.dag and the grounding-pilot mirror test that pinned them against RUST_PILOT_PRIMITIVES.
• Replaces string-based routing ("OrderedRing" / "Byte" template-name match) with declaration-identity witnesses derived from the std OrderedRing<C>/Semiring<C> declarations and TargetCarrier/IntegerAlgebra variant payload type ids — single authority is rust_pilot_primitives.
• Adds validate_rust_pilot_integer_primitives invoked from both bootstrapped Dag constructors (dag.rs:2269,2289), so corruption/drift in the pilot list surfaces at Dag::new() rather than only on a hot-path query — fail-closed.
• The previous "scaffold" comment in primitives.dag:178-181 is replaced with a comment that describes the now-actual mechanism.

Modeling-discipline alignment: facts flow forward (one source: pilot list, walked via ValueBody::List); illegal states (duplicate/missing/malformed) raise diagnostics; identity-based routing eliminates the string-matching coupling.

Minor observations (non-blocking):

• pilot_integer_row early-returns Ok(None) when payload.len() < 5, while the bootstrap validator requires < 7. Slight asymmetry, but the bootstrap gate already catches under-sized rows globally, so a query path silently treating a short row as a non-match is acceptable.
• EXPECTED_INTEGER_ROWS = 8 is a hard ratchet on pilot scope; appropriately documented as "pilot int scope" and will trip loudly when the scope grows — that's the intended behavior.
• integer_literal_routing_witness in lib.rs:1348 exposes a raw (DeclarationId, DeclarationId) tuple to integration tests rather than the IntegerRoutingWitness struct; minor leak of internal shape but contained.

[briansrls]

Re: openai-pro (gpt-5-5-pro) review on 2cf99eb4 — addressed on current tip 7cad8e2

1. P3 / missing rust_pilot_primitives: Valid for the old validate shape: let Some(pilot) = ... else { return; } could leave bootstrap with no diagnostic if the extdeps declaration was absent. Fix: that branch now attach_diagnostic with MalformedIntegerRangeFact and a SourceSpan on dsl/extdeps/languages/rust/primitives.dag (commit 7cad8e2).

2. E0502 / mutating Dag while walking pilot rows: Already fixed on this branch (not on 2cf99eb4 alone): PilotListSnapshot + ValueBody::List(els).clone() ends the pilot borrow before the row loop; the review’s line numbers referred to the pre-snapshot form. Merged in a92d35a, unchanged by this follow-up.

3. Verdict REQUEST_CHANGES: the structural direction was already correct; the remaining blocking item was the silent None return — now fail-closed.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 7cad8e23 · Trigger: schedule
• Thinking: 38s wall

---

This PR collapses the IntegerRangeFact parallel-representation bridge: integer-literal range narrowing now reads rust_pilot_primitives directly, routing by DeclarationId-equality on IntegerAlgebra/TargetCarrier variant payload types rather than template-name strings or a mirrored fact list. It also adds a fail-closed bootstrap validator and deletes the grounding-pilot mirror drift-pin.

This is exactly the "single authority / coprod dissolution / facts flow forward" move the old comment promised. Changes I checked:

• primitives.dag removes the duplicated IntegerRangeFact rows; rust_pilot_primitives becomes the only authority for bounds.
• int_literal_ranges.rs replaces String routing keys with IntegerRoutingWitness { algebra_variant_ty, carrier_variant_ty } — structural identity, not name strings.
• validate_rust_pilot_integer_primitives runs at Dag::new(), fails closed on missing pilot, missing body, wrong shape, non-parseable / inverted ranges, duplicate (algebra, carrier) witnesses, or wrong row count. Aligns with INVARIANTS' fail-closed-at-bootstrap stance.
• Test rewrites match: the duplicate/malformed-fact tests are replaced with witness-uniqueness and witness-equality tests against the pilot list — behavior-driven and hermetic.

Minor exploratory observations (not findings):

• default_span is reused for every row in the validator — diagnostics can't pinpoint the offending row within primitives.dag. Adequate at bootstrap, but a per-row span would be friendlier when the pilot list is hand-edited.
• EXPECTED_INTEGER_ROWS = 8 is a hard literal; if the pilot int scope grows, two places need editing (the .dag and this constant). Acceptable given pilot scope is locked, but worth a // dissolves when … note if this stays.
• integer_instantiation_witness still looks up "OrderedRing"/"Semiring" by name to obtain their DeclarationIds. That's fine (one resolution, then identity comparisons), just noting the boundary.

Verdict: APPROVE. Cleanly removes a parallel representation, replaces string-keyed routing with declaration-identity witnesses, and strengthens bootstrap fail-closed checks. No invariant or modeling-discipline violations observed in the diff.

[briansrls]

Re: Claude (Opus) api-review on a92d35a8 — APPROVE received; no new commit from this pass.

• Line refs: validate_rust_pilot_integer_primitives is still invoked from the two extdeps LazyLock inits in dag.rs (now ~L2235 / L2255 on tip; 2269/2289 in the review are harmless drift from column layout / commit).
• Minor notes (agreed, non-blocking): (1) pilot_integer_row <5 vs bootstrap <7: intentional split — hot path defers to bootstrap to reject malformed list rows. (2) EXPECTED_INTEGER_ROWS = 8: intentional ratchet; pilot growth will force an explicit change + regen. (3) integer_literal_routing_witness as (DeclarationId, DeclarationId): small API leak vs IntegerRoutingWitness; keep unless we need the struct in public API — out of scope for this dissolution PR.
• Tip after a92d35a: follow-up 7cad8e2 adds P3 fail-closed when rust_pilot_primitives is absent (openai review); architecture from this approve stands.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 7cad8e23 · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This PR dissolves the temporary integer-range mirror. Instead of carrying separate IntegerRangeFact declarations beside rust_pilot_primitives, integer literal narrowing now derives a routing witness from the user’s resolved std integer type, then matches that witness directly against the IntegerPrimitive rows in rust_pilot_primitives. The load-bearing move is the switch from string keys like "SemiringAlgebra"/"ByteCarrier" to declaration-identity witnesses: src/v3/compiler/src/int_literal_ranges.rs:38: /// Structural witness for routing integer literals: \IntegerAlgebra andthroughsrc/v3/compiler/src/int_literal_ranges.rs:45: pub(crate) carrier_variant_ty: DeclarationId,.

The diff also makes the new single authority fail-closed at bootstrap: both bootstrapped DAG constructors now call src/v3/compiler/src/dag.rs:2269: crate::int_literal_ranges::validate_rust_pilot_integer_primitives(&mut dag); and src/v3/compiler/src/dag.rs:2289: crate::int_literal_ranges::validate_rust_pilot_integer_primitives(&mut dag);. That validator checks that the pilot list exists, is a list, contains exactly the eight pilot integer rows, has parseable ranges, and has unique (algebra, carrier) witness pairs. Tests then move from mutating/de-duplicating the removed IntegerRangeFact authority to asserting that the pilot list itself has unique witnesses and that UInt8’s std witness matches exactly one pilot row.

2. Invariant categories

1. LAYER MODEL — Compliant. This touches substrate-adjacent extdeps data, but the direction is toward single authority: dsl/extdeps/languages/rust/primitives.dag:178: // Integer literal range narrowing walks \rust_pilot_primitives``anddsl/extdeps/languages/rust/primitives.dag:181: // witness on the user's type — single authority for bounds.replace the prior duplicate`IntegerRangeFact` bridge.
2. INVARIANTS.md + modeling-discipline.md — Compliant. Fail-closed is handled at the authoritative boundary: missing rust_pilot_primitives now attaches a diagnostic at bootstrap via src/v3/compiler/src/int_literal_ranges.rs:362: let Some(pilot) = dag.rust_pilot_primitives() else { and src/v3/compiler/src/int_literal_ranges.rs:363: dag.attach_diagnostic(malformed_integer_range_fact(, while duplicate witness pairs are rejected by src/v3/compiler/src/int_literal_ranges.rs:569: if !witnesses.insert((*algebra_ctor, *carrier_ctor)) {.
3. CODING.md — Finding, NON-BLOCKING. src/v3/compiler/src/int_literal_ranges.rs:348: pub(crate) fn validate_rust_pilot_integer_primitives(dag: &mut Dag) { is a newly added large validator spanning several separable jobs: resolving schema constructors, snapshotting the pilot list, validating row shape, parsing numeric bounds, checking duplicate witnesses, and checking total row count. That is implementation-only and not a correctness blocker, but it does strain the small/composable function guideline; a follow-up split into “schema resolution,” “row validation,” and “aggregate validation” helpers would make the boundary easier to audit.
4. TESTING.md — Compliant. The tests are aimed at the changed behavior rather than the deleted mirror: src/v3/compiler/tests/integration/int_literal_cardinality_test.rs:327: fn rust_pilot_primitives_integer_witnesses_are_unique() { pins uniqueness of pilot witnesses, and src/v3/compiler/tests/integration/int_literal_cardinality_test.rs:389: fn int_literal_range_routing_matches_std_type_witness() { checks the std type witness against the pilot rows with a minimal UInt8 fixture.
5. LOCKED DESIGN DECISIONS — N/A. I do not see this diff referencing or altering a locked thesis/design decision; the direction is consistent with the locked-style single-authority discipline rather than diverging from it.
6. TRACKED vs UNTRACKED DEBT — Compliant. No new scaffold/TODO bridge is introduced; the prior scaffold is dissolved, and the replacement comment explicitly names rust_pilot_primitives as the authority at dsl/extdeps/languages/rust/primitives.dag:178: // Integer literal range narrowing walks \rust_pilot_primitives``.

3. Verdict

APPROVE_WITH_COMMENTS

The substrate/modeling direction is sound: the duplicate integer-range authority is removed, the pilot list becomes the source of truth, and bootstrap validation fails closed when that authority is malformed. My only comment is implementation-style, not semantic: the new bootstrap validator is large enough that factoring it would make future audits cheaper.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 7cad8e23 · Trigger: schedule
• Thinking: 61s wall

---

✅ No blocking issues found; the migration to declaration-identity witness matching plus bootstrap-time rust_pilot_primitives validation is coherent, increases fail-closed coverage, and removes the redundant IntegerRangeFact bridge cleanly.

[briansrls]

Director — stale, needs worker refresh.

This is Cleanup Worker 3 / vivid-carp-264 (target primitive grounding / string-routing dissolution). Last updated 12:42Z (~12+ hours ago), DIRTY, body still dashboard-default, title still slug.

Per earlier Cleanup sweep, blocked on:

• fmt red
• v3 regen_bootstrap --verify stale bootstrap snapshots

Status check:

• Has vivid-carp-264 surfaced any progress today, or is the worker silently stalled?
• 6 files +592/-389 is substantive scope (target primitive grounding + new int_literal_ranges.rs); the substance looks promising

Cleanup Manager (#941): if vivid-carp-264 is silent, redispatch to a fresh worker. The work scope is sound but the implementation needs:

1. Refresh against current main (post all the recent merges — feat(v3): add ValueBody map carrier #1017, gentle-badger-838 #1063, sleek-wren-716 #1004, etc.)
2. Re-run regen_bootstrap --verify to sync bootstrap snapshots
3. Fill PR body with proper scope statement + per-PR-gate (b) disposition

Don't merge until refresh + body land. The dashboard auto-attach event may have just re-pushed the stale state.

[briansrls]

#1005 refresh is now pushed at f9923f8d.

• Merged current origin/main and resolved the grounding_pilot conflict by keeping main's Go range-fact drift checks while preserving this PR's Rust IntegerRangeFact mirror deletion.
• Ran cargo fmt --all; pre-push also re-ran cargo fmt --all --check.
• Regenerated bootstrap snapshots and verified regen_bootstrap --verify.
• Updated the PR title/body with scope, gate disposition, and verification.
• Local checks passed: cargo fmt --all --check, cargo run -p v3-compiler --features bootstrap-regen-fresh --bin regen_bootstrap -- --verify, cargo build -p execute-command-bootstrap, and RUSTC_BOOTSTRAP=1 cargo test -p v3-compiler -- -Z unstable-options --report-time (594 passed, 0 failed, 25 ignored).

GitHub now reports mergeable: MERGEABLE; checks are queued on the refreshed head.

— sent from tidy-dove-734

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: f9923f8d · Trigger: schedule
• Thinking: 46s wall

---

This PR cleanly executes a coprod-dissolution: removes the IntegerRangeFact mirror, rewrites the consumer to walk rust_pilot_primitives via declaration-identity witnesses (no string routing on template/variant labels), and adds a fail-closed bootstrap validator with a structural drift test.

Findings: none worth flagging.

• Single-authority restored: pilot rows deleted (primitives.dag:178-275 deletion block), comment updated to declare authority (primitives.dag:178-181).
• Routing now uses DeclarationId equality on OrderedRing/Semiring templates and Byte/Word{16,32,64} carriers (int_literal_ranges.rs integer_instantiation_witness / std_word_carrier_to_target_carrier_variant_ty) — eliminates the prior name-string lookups; consistent with INVARIANTS "facts flow forward" / API-level enforcement.
• Bootstrap gate validate_rust_pilot_integer_primitives runs at Dag::new() (dag.rs:2321,2339), checks structure, parses to i128, enforces min ≤ max and witness uniqueness, fails closed on missing pilot — appropriate fail-closed posture.
• The grounding-pilot drift test deletion (grounding_pilot/src/lib.rs) is correct since the bridge it pinned no longer exists.
• EXPECTED_INTEGER_ROWS = 8 is a bounded scaffold tied to the documented "pilot int scope" — fine; will scream if drift occurs.

Exploratory observation (non-blocking): validate_rust_pilot_integer_primitives clones the entire ValueBody::List into PilotListSnapshot::List(els) to dodge the &dag / &mut dag borrow conflict for attach_diagnostic. Workable, but if pilot lists grow this borrow pattern will likely recur — could later be cleaner with a "collect diagnostics, attach after" split.

Verdict: APPROVE — diff faithfully dissolves the duplicate substrate, replaces text-keyed routing with identity witnesses, and adds an appropriate fail-closed bootstrap gate. No invariant violations observed.

[briansrls]

Review at f9923f8d has no actionable findings to address. I rechecked the live gate: GitHub reports mergeable: MERGEABLE; fmt, ci, and v3 are green, with self_host_ratchet still in progress. No code change needed for this review item.

— sent from tidy-dove-734