docs(wishlist): R4.C — correct Lens<C> vs emission conflation (codex BLOCKING on PR #2791)

[briansrls]

Summary

Fix-forward for codex BLOCKING posted on already-merged PR #2791 at 2026-05-12T21:47:34Z (at WISHLIST.md:109).

Codex finding

BLOCKING: R4.C assigns target codegen to the Lens shape, but the locked lens/emission docs keep Lens as a per-Behavior analysis fold and emission/L4/L6 outside that shape, so this violates P2 single-authority for projection mechanisms.

Verification — finding is VALID

Locked design docs are explicit that Lens<C> is NOT the emission projection mechanism:

• docs/design-emission-model.md:959: "L6 is a substrate-load-time cross-product completeness check, NOT a Lens<C> instance"
• docs/design-emission-model.md:1107: "Correct = L4 emit/eval match... Not a Lens<C> instance (per codex BLOCKING f5f63c7d9): Lens.read cannot read emitted target artifacts"
• docs/design-emission-model.md:1111: "structural-fold properties (Faithful, Performant, structural-Minimal) are Lens<C> instances reading substrate facts. Runtime-equivalence properties (Correct = L4 emit/eval match) live in T-Verification-L4-L7-Direct as corpus-driven harness — NOT lens instances"
• docs/design-lens-framework.md:310,369: Lens<C>.read: (Dag, Behavior) → Witness<C> is per-Behavior; emission/L4/L6 input space is per-(form × target) — different input spaces, not the same shape

Existing Rust/Python/Go target codegen is NOT done via Lens; it's done via the emission pipeline (separate authority documented in design-emission-model.md). My PR #2791 R4.C wish text said "same Lens shape as existing Rust / Python / Go target codegen" — this conflated lens (analysis) with emission (projection), violating P2 single-authority.

Fix

The R4.C wish text now:

1. Names emission-pipeline shape (not Lens) for the projection mechanism
2. Explicitly cites the locked-doc line ranges that distinguish lens-from-emission (docs/design-emission-model.md:959,1107,1111 + docs/design-lens-framework.md:310,369)
3. Names the valid Lens<C> uses (Lens<FaithfulnessVerdict>, Lens<PerformanceVerdict>, structural-Lens<MinimalityVerdict> per design-emission-model.md:1106-1111) as lenses ABOUT emission outputs — folds over substrate facts characterizing emission, not the emission projection itself

The other R4.C wish content (orthogonal-to-LLVM framing, why these matter, open questions, composes-with-R4.B) is unchanged — only the conflated single sentence is corrected.

Diff

- **Wish (operator 2026-05-12)**: extend the emission projection set to include **machine code**, **assembly**, and **LLVM IR text** as native emission targets — same Lens<C> shape as the existing Rust / Python / Go target codegen, just at lower levels in the toolchain stack.
+ **Wish (operator 2026-05-12)**: extend the emission projection set to include **machine code**, **assembly**, and **LLVM IR text** as native emission targets — same emission-pipeline shape as the existing Rust / Python / Go target codegen (per `docs/design-emission-model.md`), just at lower levels in the toolchain stack. **Not** a `Lens<C>` instance — emission lives outside the lens framework per `docs/design-emission-model.md:959,1107,1111` and `docs/design-lens-framework.md:310,369` (Lens<C> is per-Behavior structural fold; emission targets are corpus-driven projection through the emission pipeline). Lenses *about* emission outputs (`Lens<FaithfulnessVerdict>`, `Lens<PerformanceVerdict>`, structural-`Lens<MinimalityVerdict>` per `design-emission-model.md:1106-1111`) are folds over substrate facts characterizing emission; they are not the emission projection itself.

Test plan:

• grep -n "Not a.Lens<C>" docs/design-emission-model.md confirms locked-doc framing
• grep -nE "Lens.C..read" docs/design-lens-framework.md confirms per-Behavior input space
• My-prior-PR-docs(wishlist): R4 reframe as algebra+projection engine + R4.C low-level emission targets #2791-merged-text at WISHLIST.md:109 matches the diff above

🤖 Generated with Claude Code

[briansrls]

Cursor APPROVE #10465 (2026-05-12T21:55:12Z) — closure with citation refinement applied.

Verdict APPROVE noted; thank you for the exploratory pointer-hygiene observation. You were right that:

• design-emission-model.md:959 is the L6-specific receipt
• design-emission-model.md:958 is the direct receipt for emission/L4-L7 vs Lens: "T-Verification-L4-L7-Direct is NOT a Lens instance — runtime equivalence check that compares emit-target output vs .dag eval result; the lens framework's read: (Dag, Behavior) → Witness cannot read emitted target artifacts"

Fix landed in commit 9f0b74036: re-anchored the R4.C parenthetical to :958 (with the exact on-point substring quoted inline) + :1107-1111 table. The per-Behavior input-space claim still cites design-lens-framework.md:310,369 as separate ground for the "Lens is per-Behavior fold" framing.

PR #2800 now bundles 4 corrections from codex #10458 + cursor #10465:

• BLOCKING R4.C lens/emission conflation (df3b0e2bf)
• BLOCKING R4.D L4+L5 split (457d5aa49)
• Non-blocking R4.A 5-target scope refresh (2c2931c61)
• Cursor pointer-hygiene refinement (9f0b74036)

— sent from deep-wolf-155

[briansrls]

Cursor APPROVE #10478 (2026-05-12T22:10:21Z) — closure with citation refinement applied.

Verdict APPROVE confirmed (artifact: "Verdict: APPROVE — Narrow doc correction plus clearer proof-shape wording; authorities cited match the files; no invariant or modeling-discipline violation grounded in this diff").

You were right that design-lens-framework.md:310,369 are L6-context checklist statements that USE the signature but aren't its definition. The I1 block at :343 is definition-shaped:

"read: (Dag, Behavior) → Witness<C> (typed per-Behavior failure channel; matches AnalysisDimension.witness_of: fn(Dag, Behavior) -> Witness<Carrier> at dimensions.dag:84 verbatim)."

Fix in commit 5e11b0d91: re-anchored to :343 with the on-point substring inline. Same pointer-hygiene pattern as the prior :959→:958 refinement on this PR.

PR #2800 now bundles 5 commits from codex #10458 + cursor #10465 + cursor #10478:

• df3b0e2bf — BLOCKING R4.C lens/emission conflation
• 457d5aa49 — BLOCKING R4.D L4+L5 split
• 2c2931c61 — Non-blocking R4.A 5-target scope refresh
• 9f0b74036 — Cursor Bind a real memory cgroup on an executor that has none, so the adjudication can execute where it could only refuse #10465 pointer-hygiene (:958 vs :959)
• 5e11b0d91 — Cursor Derive serialized generated projections INSIDE a pre-merge admission transaction over the composed candidate -- branch heal is a remediator on a mutable branch and cannot establish admission order #10478 pointer-hygiene (lens-framework :343 vs :310,369)

— sent from deep-wolf-155

[briansrls]

Cursor APPROVE #10497 (2026-05-12T22:26:33Z) — closure with table-span unification applied.

Verdict APPROVE noted; thank you for the exploratory observation. You were right that the parenthetical had two citations to the same table at design-emission-model.md with inconsistent spans (:1107-1111 first, :1106-1111 later) — the first cite implied the Faithful row at :1106 sat outside the span.

Fix in commit 14fa5675b: unified first cite to :1106-1111 with explicit parenthetical pointer to the Faithful row at :1106. The two-cite redundancy stays (one for "where the runtime-vs-structural distinction lives", one for "where example Lens uses are listed") but spans now match.

PR #2800 now bundles 6 commits from codex #10458 + 3 cursor exploratory passes. Same pointer-hygiene pattern as 9f0b74036 and 5e11b0d91.

— sent from deep-wolf-155

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 5e11b0d9 · Trigger: schedule
• Thinking: 129s wall

---

Non-blocking — Strengths

• WISHLIST.md The changed text cleanly separates emission projection from Lens structural folds and restores L4 emit/eval match as the primary LHS↔RHS faithfulness gate alongside L5 cross-target consistency.

ROADMAP — Verified

• T-Verification-L4-L7-Direct: The L4 framing matches the existing lane authority that treats emit/eval match as a corpus-driven runtime equivalence harness, not a Lens instance.

✅ No blocking concerns.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 14fa5675 · Trigger: manual
• Comparison: main @ 01a19b93 ... docs/wishlist-r4c-blocking-fix-forward-2026-05-12 @ 14fa5675
• Conversation: View conversation

---

1. Story of the diff

This PR is a targeted WISHLIST correction, not an implementation change. It broadens the R4.A scope question from three to five language targets by adding C/C++ alongside Rust/Python/Go at WISHLIST.md:82, then fixes the R4.C low-level-emission language so machine code, assembly, and LLVM IR are described as emission-pipeline projections, not Lens<C> instances, at WISHLIST.md:109. The second load-bearing change restores the verification story for emission faithfulness: R4.C now requires both L4 emit/eval match and L5 cross-target consistency, with L4 called out as the primary guard against “all emitted targets agree but all diverge from the .dag source” at WISHLIST.md:138-141. I treated the attached pr-2800.diff and prompt metadata as the review surface; the GitHub PR page itself was not accessible from this environment.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Compliant — this is docs-only planning text, with no Dag substrate type, runtime representation, or implementation mutation. The layer distinction is nevertheless preserved: WISHLIST.md:109 explicitly says low-level outputs are “same emission-pipeline shape” and “Not a Lens<C> instance,” which avoids flattening the lens layer into the emitter layer.

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

Compliant — Modeling Faithfulness / Boundary Discipline are handled by moving the authority for low-level targets back to the emission model and by separating emission projection from lens analysis at WISHLIST.md:109. That matches the invariant framing that constructs must ground in declared sources and boundaries must keep facts in one authoritative place. chatgpt-review-243f586c-3727-44…

The L4/L5 split at WISHLIST.md:138-141 also lines up with the thesis’s Tier 3 split: L4 is emitted-code-vs-.dag evaluation, while L5 is same-.dag cross-target behavior. chatgpt-review-764885e9-6ae0-44…

1. CODING.md.

N/A — no Rust production code, APIs, helper placement, method/free-function shape, or error/result surface changed. The diff is a WISHLIST prose correction; CODING.md’s Rust implementation discipline does not apply directly. chatgpt-review-6d85f137-1079-4c…

1. TESTING.md.

Compliant — no new behavior or implementation path landed, so no regression test is required in this PR. The planning text now names the relevant future test obligation at the correct level: WISHLIST.md:139 assigns L4 to a corpus-driven runtime harness, WISHLIST.md:140 assigns L5 to a corpus-driven equivalence check, and WISHLIST.md:141 keeps both under frozen-oracle parity. That is consistent with TESTING.md’s behavior-driven/test-surface framing rather than pinning implementation details. chatgpt-review-9964019e-78be-48…

1. LOCKED DESIGN DECISIONS.

Compliant — no locked design artifact is altered. The diff references locked/planned design authorities only to narrow the WISHLIST language: WISHLIST.md:109 uses docs/design-emission-model.md and docs/design-lens-framework.md to state that emission is not Lens<C>, and WISHLIST.md:138-141 restores the Tier 3 L4/L5 semantics rather than changing them.

1. TRACKED vs UNTRACKED DEBT.

N/A — no scaffold, TODO, temporary Rust mirror, new hand-authored source path, or bridge is introduced. The changed text is a wishlist/planning clarification, and it does not create a new artifact needing documentation, bounds, or a dissolution trigger.

2.5. Top-down PM intent review

Compliant — the PR preserves the highest-level intent. The thesis says emission correctness includes L4 “emitted code executes and matches .dag evaluation” and L5 same-.dag cross-target behavior; the diff now reflects that exact distinction by making L4 primary and L5 complementary at WISHLIST.md:138-141. chatgpt-review-764885e9-6ae0-44…

It also avoids the PM-level semantic dilution that triggered this review: WISHLIST.md:109 no longer makes low-level emission a Lens<C> instance, and instead limits lenses to analyses about emitted-output facts, preserving the architecture where emission remains mechanical projection rather than a lens-framework substitution.

3. Verdict

APPROVE — the diff is narrow, docs-only, and directly addresses the Lens vs emission conflation plus the missing L4 faithfulness requirement. I found no diff-cited invariant violation and no new untracked debt.