TypeScript RCA Manager — per #4137 §11.8 ALPHA/PREVIEW lane (NOT release-minimum per Wave F F3). Currently no worksheets started. Goal by end of next iteration: L0 complete per #4117 pattern. Author worksheets for TS R2a/R2b/R3-external leaf-model widening, TS TargetAtomRealization rows, TS TargetTy

[briansrls]

Auto-opened by session-dashboard for session fierce-fox-719.
Pushing to session/fierce-fox-719 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]

Review metadata

• Provider / model: codex / unknown
• Commit: e63b8cbd · Trigger: schedule
• Thinking: 315s wall

---

BLOCKING (2)

Root Cause

• docs/planning/v4-ts-leaf-model-r2-r3-external-worksheet-2026-06-01.md R2b uses a rounded power-of-two boundary as the number-lane negative witness → replace it with a checkable precision-loss fixture whose number verdict is distinct from the exact bigint model.
• docs/planning/v4-ts-target-atom-realization-worksheet-2026-06-01.md TS needs a Symbol factory value realization but the plan assumes the existing SG-1 carrier can already express factory calls → ratify the missing std target_model value-template shape before dispatching the TS row worker.

⚠️ Two worksheet entries would send downstream implementation into an unverifiable R2b probe and an unrepresentable TS Symbol realization.

[briansrls]

BLOCKING: The R2b falsification is not falsifying: ((2 ** 63 - 1) + 1) === 2 ** 63 evaluates true under JavaScript number rounding, so the wrong-path probe can pass with the same verdict as bigint, violating the leaf-model falsification contract and INVARIANTS P1.

[briansrls]

BLOCKING: value_form: Symbol(<string-literal>) is not representable by the landed TargetAtomRealization/TargetValueTemplate vocabulary, so the worksheet either requires changing the std carrier it says to consume or creates a parallel TS value-template authority, violating INVARIANTS P2.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: e63b8cbd · Trigger: manual
• Comparison: main @ 92675275 ... session/fierce-fox-719 @ e63b8cbd
• Conversation: View conversation

---

1. Story of the diff

This PR adds five planning worksheets for the TypeScript alpha/preview lane, without changing compiler or .dag implementation code. The worksheets carve the L0 path into prerequisite substrate/modeling work: expose stable TypeScript algebra inhabitance fact IDs, add TS TargetTypeExpressionProjection, use that projection from TS TargetAtomRealization rows, and then consume those facts from TS R2a/R2b/R3-external leaf-model claims and host runners. In parallel, it adds a TS grammar-inverse TestClaims worksheet scoped to a small G1/G2 production set and explicitly keeps packaging/module resolution out of L0. The load-bearing shape is mostly single-authority and staged: shared carriers stay in std/, TypeScript rows land in src/v4/extdeps/languages/typescript.dag, claims reference fact IDs rather than new Subject/Expectation coproducts, and implementation workers are blocked on Modeling DFS Arbiter approval.

2. Invariant categories

1. LAYER MODEL — Compliant. The worksheets are planning-only, but the planned substrate split is correct: shared carriers are consumed from src/v4/std/target_model.dag, TS-specific rows are placed in src/v4/extdeps/languages/typescript.dag, and the algebra worksheet rejects new Phase-1 Subject/Expectation coproducts in favor of LeafModelClaimId plus stable fact Symbols.
2. INVARIANTS.md + modeling-discipline.md — Finding. docs/planning/v4-ts-leaf-model-r2-r3-external-worksheet-2026-06-01.md:90 says “This worksheet PR (docs(planning): TypeScript L0 §10.0 worksheets (#4137 §11.8 alpha lane) #4147) lands markdown only,” but this review surface is PR TypeScript RCA Manager — per #4137 §11.8 ALPHA/PREVIEW lane (NOT release-minimum per Wave F F3). Currently no worksheets started. Goal by end of next iteration: L0 complete per #4117 pattern. Author worksheets for TS R2a/R2b/R3-external leaf-model widening, TS TargetAtomRealization rows, TS TargetTy #4206. For a normal prose doc that might be a typo; here it is inside the §P5(b) receipt section that downstream workers will use for SG-0/P5 traceability. INVARIANTS P5 says scaffolds/receipts need checkable completion conditions and traceable dissolution; a stale PR number makes the receipt point at the wrong authority and should be corrected before this planning artifact becomes the worker contract.
3. CODING.md — N/A. The diff adds Markdown planning worksheets only; there is no new Rust implementation surface, helper placement, method shape, or comment-vs-structure issue to review under CODING.md.
4. TESTING.md — Finding. docs/planning/v4-ts-leaf-model-r2-r3-external-worksheet-2026-06-01.md:67-68 says the R2b falsification “uses number lane for same magnitude and expects IEEE754 precision loss or tsc rejection,” and :83 reduces that to “Same expr in number lane → diverges from model.” The happy expression is (2n ** 63n - 1n) + 1n === 2n ** 63n; the “same expr” in number form does not name a concrete failing predicate and can be implemented as an equality that passes for the wrong reason. For L0, the worksheet must pin an observable negative behavior, for example a number-lane assertion that exact successor is preserved beyond the safe-integer boundary and is expected to fail under JS number, rather than leaving “precision loss or tsc rejection” as an implementation choice. chatgpt-review-02c6d3c0-af72-40…

TESTING.md — Finding. docs/planning/v4-ts-leaf-model-r2-r3-external-worksheet-2026-06-01.md:116 specifies the typecheck authority as tsc --strict or npx tsc with a repo-pinned TS version, while the same worksheet’s fixtures rely on 2n bigint literals and Symbol("x") value availability at :83-84. That is not hermetic enough for a host-runner acceptance contract: the worksheet should pin the full TS compile surface needed by these fixtures, such as the intended --target, --lib, module/no-emit behavior, and whether a local tsconfig is forbidden or supplied. Otherwise a downstream worker can get environment-dependent failures unrelated to the modeled TS facts.

1. LOCKED DESIGN DECISIONS — Compliant. I do not see a locked-design divergence in the changed lines. The docs repeatedly preserve the canonical shared-carrier home, forbid redefinition of TargetAtomRealization / TargetTypeExpressionProjection, keep Shape-A TS target modeling in extdeps, and explicitly mark the lane as alpha/preview rather than a release-minimum gate.
2. TRACKED vs UNTRACKED DEBT — Compliant aside from the stale PR-number finding above. The PR introduces several bridges or staged items, but the important ones are bounded: host runners are marked interim, the hand-Rust boundary receipt names the INVARIANTS row, census increment, verify shells, and T-22 dissolution trigger, and the SG-2 ProjectionAbsent shim has a named removal trigger once the TS row is present.

2.5. Top-down PM intent review

Finding. The PM intent is L0 completion per the #4117-style pattern: the worksheet itself states acceptance must be “dual-path verification (happy + falsification) against tsc and Node,” not compile-only wiring. But the R2b row then gives a non-specific “Same expr in number lane → diverges from model” falsification, which can guide a faithful worker to author a negative test that does not actually falsify exact integer behavior. That dilutes the highest-level intent from “L0 complete with real falsification” into “happy path plus a vaguely expected numeric discrepancy.” chatgpt-review-02c6d3c0-af72-40…

3. Verdict

REQUEST_CHANGES. The worksheet structure is directionally strong and mostly respects single-authority modeling, but the R2b falsification and TypeScript toolchain contract are not precise enough to authorize L0 implementation work, and the §P5(b) receipt carries a stale PR number. Fixing those keeps the planning artifact faithful before downstream workers execute it.

[briansrls]

Closing as SUPERSEDED. Jun-1 TS-RCA planning worksheets (leaf-model/atom-realization/type-expr-projection/grammar-inverse). Superseded by ctrl#1489 (emit-ladder authority); TS type-expr work is folded into the T3 type-tier lane under Mgr-SPINE. Operator-authorized close (via still-raven-546).