docs(roadmap): 3 novel debt rows + pipeline_authority amendment post-#637

[briansrls]

Summary

Three novel debt rows surfaced by the 2026-04-22 exploratory analysis (gpt-5-4-pro against main@c9758c8), each verified against HEAD before adding. Plus amendment of the pipeline_authority row to reflect #637's partial close.

Novel (3 new rows)

1. __EMIT_BUG_{0}__ fabrication sentinels across 3 target emit specs — dag/emit.dag:54, go/emit.dag:107, python/emit.dag:115 use sentinel-string fabrication; rust correctly uses compile_error!(). Target behavior already diverges between fail-closed and fabricated output. P3 / C-6 violation. S scope.

2. algebra.dag declaration-vs-template parallel authority — OrderedRing<T> declaration (16 fields) vs ordered_ring_templates() (9+clamp) already drifted. Same pattern for PartialFunction. Distinct from already-tracked signature/comment-mismatch debt — that's prose vs declaration; this is code vs declaration. P2. M scope.

3. browser.dag typed-carrier service-boundary collapse — typed BrowserConfig / BrowserContext / Page / Element / Url declared; service ops use raw String (headless, url, selector, ...). Same class as tracked LLM + GitHub rows. P1 / M8. M per service family.

Amendment (1 existing row)

pipeline_authority.rs — #637 closed the stage-bindings portion (structural PipelineStageBinding consumption landed). Residual: pipeline_compile_order_names() at :225-275 still text-slices the compile-fn body for stage order. Blocked on substrate capability for ordered-body reflection (related to SG-2c-proper emit-capability program). Named the trigger explicitly.

Not a new row

ResourceHandle forgeability (also cited by the analysis) verified as already tracked at ROADMAP.md:138. No action.

Test plan

• Verify the three new claims are accurate (spot-check each file:line citation)
• Confirm the pipeline_authority amendment reads accurately against current pipeline_authority.rs:225-275 post-Pipeline Authority — structural-read cleanup #637

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: codex / gpt-5.4
• Commit: 457da80c · Trigger: schedule
• Thinking: 134s wall

---

Findings

• ROADMAP.md:180 violates INVARIANTS.md's P1 "Documentation Describes Live State" and P2 single-authority discipline. The new text says the remaining stage-order fact "lives in the compile fn's procedural body" and is blocked on ordered-body reflection, but the landed code in src/v3/compiler/src/pipeline_authority.rs already treats PipelineStageBinding declaration order as the runtime authority and uses the compile body only as a fail-closed drift check. Reframing this as a substrate-capability blocker misstates where the authority actually lives.

Verdict
REQUEST_CHANGES. The three new debt rows read cleanly, but the post-#637 pipeline_authority amendment does not match the implementation's current authority model, so the roadmap entry should be corrected before merge.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 457da80c · Trigger: schedule
• Thinking: 137s wall

---

Findings

Non-blocking — amended pipeline_authority row has wrong line refs / misframes the residual (ROADMAP.md:180):

The amendment says "pipeline_compile_order_names() at :225-275 still include_str!s pipeline.dag, slices the compile-fn body span, and iterates body.lines() for stage order."

Against the actual post-#637 pipeline_authority.rs (311 lines, verified in-worktree):

• pipeline_compile_order_names() is at :185, and is a thin ordered_pipeline_stages(dag).map(.stage_name) wrapper — it does no text slicing.
• The residual text-slicing lives in compile_body_stage_names_from_source() at :130-180, called only from reconcile_with_compile_body() at :110-128.
• Crucially, reconcile_with_compile_body is a fail-closed cross-check that errors when the fn compile body drifts from the PipelineStageBinding declaration order. The bindings are already the runtime authority post-Pipeline Authority — structural-read cleanup #637; the slicing is a defensive drift detector, not a "still reading stage order from source text" residual.

So both the line range and the framing are off. A future reader grepping :225-275 won't find the residual, and "reads stage order from source text" understates what #637 actually landed. Suggested rewrite: point at compile_body_stage_names_from_source() at :130-180 and describe the residual as "fail-closed compile-body cross-check still text-slices; converting to structural ordered-body reflection needs substrate capability X."

Tracked-debt hygiene point, not a principle violation — the dissolution trigger is documented, bounded to "substrate gains ordered-body reflection," and named. Fix in a follow-up or a light amend.

Verdict

APPROVE_WITH_COMMENTS — docs-only diff, small and narrowly scoped. The 3 new debt rows (EMIT_BUG sentinel across dag/go/python emit specs, algebra.dag declaration-vs-*_templates() parallel authority, browser.dag typed-carrier collapse at the service boundary) all check out against the cited code. The pipeline_authority amendment is useful in intent — correctly noting that #637 closed the bindings portion — but the residual pointer is inaccurate enough that I'd fix it before the row loses traceability value.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-4-pro
• Commit: 457da80c · Trigger: manual
• Conversation: View conversation

---

1. Story of the diff

This is a docs-only ROADMAP.md update that does two things. First, it amends the existing pipeline_authority debt entry so it no longer overclaims what #637 solved: ROADMAP.md:180 now says the stage-binding portion was fixed structurally, while the remaining debt is specifically the compile-stage order still being reconstructed from raw pipeline.dag body text and therefore blocked on ordered-body reflection. Second, it adds three newly surfaced debt rows at ROADMAP.md:196, ROADMAP.md:198, and ROADMAP.md:200 for fail-open emit sentinels in target specs, declaration-vs-template duplicate authority in algebra.dag, and typed carriers collapsing back to strings at the browser service boundary. The through-line is good: make latent invariant violations visible in the roadmap, and sharpen an older debt row so it points at the actual missing substrate capability instead of a vaguer “cleanup residual.”

2. Invariant categories

1. LAYER MODEL (substrate vs implementation). N/A — the diff is pure roadmap prose; it does not change substrate types, Dag-resident facts, or implementation code.
2. INVARIANTS.md + modeling-discipline.md. Compliant — ROADMAP.md:180 now names the real remaining authority problem and a concrete dissolution condition: “Dissolution trigger: when .dag authority gains ordered-body reflection ... replace text-slicing with typed read.” That matches the repo’s single-authority / typed-query discipline and its requirement that transitional shapes carry explicit dissolution paths (INVARIANTS.md:98-118, INVARIANTS.md:221-252).
3. CODING.md. N/A — CODING.md is scoped to Rust implementation style in src/v3/compiler/src/, and this PR changes only roadmap tracking prose, not Rust code (CODING.md:8-27).
4. TESTING.md. N/A — this PR does not change runtime behavior, claim behavioral parity, or promote any capability row; it only records debt in ROADMAP.md, so there is no same-PR test obligation under the current testing doc (TESTING.md:10-24).
5. LOCKED DESIGN DECISIONS. N/A — the diff does not alter a locked thesis/design mechanism; it only updates how existing debt is described and queued.
6. TRACKED vs UNTRACKED DEBT. Finding — ROADMAP.md:196 says “Dissolution: converge the three target emit specs on their language's native compile-time error form”; ROADMAP.md:198 says “Dissolution: *_templates() functions should be derived from the type declaration”; ROADMAP.md:200 says “Dissolution: route service ops through the declared carriers.” Those are good desired end states, but unlike ROADMAP.md:180 they do not name a specific, checkable dissolution trigger. The repo’s current debt discipline is explicit that transitional / temporary shapes need a named trigger, otherwise the debt is documented but still under-specified as a tracked receipt (INVARIANTS.md:233-252; modeling-discipline.md:95-106).

3. Verdict

APPROVE_WITH_COMMENTS

The pipeline_authority amendment is materially better than the pre-#637 wording, and the three new rows are useful debt surfacing. My only real comment is that the new entries at ROADMAP.md:196, :198, and :200 stop at problem statement + scope and do not yet meet the repo’s stronger “named dissolution trigger” standard, so they are helpful but not fully tracked in the same way as ROADMAP.md:180.

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 7362672f · Trigger: schedule
• Thinking: 37s wall

---

Verdict: APPROVE — documentation-only change to ROADMAP.md; all claims spot-check clean.

Spot-checked:

• pipeline_authority.rs amendment is accurate: ordered_pipeline_stages() is structural, pipeline_compile_order_names() at line 184 is a thin wrapper, and compile_body_stage_names_from_source() is invoked only from reconcile_with_compile_body() as a fail-closed drift check (exactly as the row describes).
• __EMIT_BUG_{0}__ sentinels present in dag/emit.dag:54, go/emit.dag:107, python/emit.dag:115; Rust uses compile_error!("{0}") at rust/emit.dag:141. Divergence claim holds.
• All new rows include file:line anchors, bounded scope, and dissolution triggers — tracked-debt rather than untracked-debt shape.

The "RESOLVED" row self-corrects an earlier (incorrect) review characterization and is retained as a historical cross-reference with an explicit suggestion to sweep it in the stale-receipt pass — appropriate handling.

No findings against INVARIANTS / modeling-discipline / CODING / TESTING — the diff doesn't touch code, and the debt rows it adds align with P1/P2/P3 framing as used elsewhere in the file.

[briansrls]

Review metadata

• Provider / model: codex / gpt-5.4
• Commit: 7362672f · Trigger: schedule
• Thinking: 97s wall

---

APPROVE

Diff is docs-only and looks clean. I spot-checked the amended pipeline_authority row against src/v3/compiler/src/pipeline_authority.rs and the three new debt rows against the cited files; the roadmap text matches the live tree and keeps the debt tracked with concrete dissolution paths, so I don’t see a violation of the pinned invariants/modeling/coding/testing rubric in this PR.

[briansrls]

Review metadata

• Provider / model: codex / gpt-5.4
• Commit: 7362672f · Trigger: schedule
• Thinking: 426s wall

---

Non-blocking — Improvements (fix in-PR if easy, else defer to roadmap)

• ROADMAP.md The new EMIT_BUG{0}_ row contradicts Lane F's own receipt: PR #657 says Python is not fail-closed under deferred annotations, so the roadmap should not say all four targets already fail closed.

ROADMAP — Verified

• PR classification: This is a design/docs-only PR because the GitHub diff changes only ROADMAP.md.
• pipeline_authority structural read: ordered_pipeline_stages() and PR #637 do make PipelineStageBinding declaration order the authority, with compile-body text slicing retained only as a fail-closed drift check.
• algebra declaration-vs-template asymmetry: dsl/std/algebra.dag really does split authority today: declarations carry operator-facing Arrow fields while *_templates() carry per-method metadata and kernel synonym surfaces.
• browser typed-carrier boundary debt: dsl/extdeps/browser.dag declares BrowserConfig/BrowserContext/Page/Element and imports Url while several service inputs and outputs still use raw String, so the debt itself is grounded.

✅ The roadmap update is otherwise aligned with the checked code and thesis/invariant framing; I only see the one wording fix above.