v4: cut work-direction meta-layer + node.dag Practice-4 ledgers (operator-ratified)

[briansrls]

Summary

Two operator-ratified (2026-05-15) v4 substrate-hygiene changes, one PR, two commits. Operator-directed cleanup — please review/merge personally; not auto-merging.

Commit 1 — cut the work-direction meta-layer

The v3-derived "model our own work-direction as .dag data" substrate is not used by the project, and the compiler model self-justifies (rationale emergent from composition), so no meta-layer narrates it. Deleted (never filled — scaffolds only; nothing in the substrate imported them, verified — pure scope reduction):

src/v4/workflow/{worker_output,brief,cycle,retirement,doc_anchor}.dag

Kept: workflow/{bootstrap,ci}.dag — compiler build infrastructure, explicitly NOT the meta-layer. bootstrap.dag is load-bearing for the anti-regression guarantee (STRUCTURE.md invariant 7); cutting it would reintroduce the v3 regression door — so it stays.

Authority docs narrowed surgically (not gutted), dated + honest per the thesis's own no-fabrication discipline:

• STRUCTURE.md — workflow/ tree → {bootstrap,ci}; checksum 68→63 .dag; invariant 6 "Workflow substrate first" → "Meta-process as data" (work-direction-substrate claim retracted; lens-self-application to build/CI survives).
• THESIS.md — facet 4 narrowed: keeps lens self-application to the build/CI pipeline (bootstrap/ci), retracts the T-Workflow-As-Data process-self-modeling sub-claim. Meta-process bullet narrowed.
• TASKS.md — T-5 → tombstone (decision on record, not silently dropped); bootstrap=T-20, ci=T-24 stand alone.
• BRIEF_TEMPLATE.md — DEFINITION-OF-DONE "WorkerOutput instance" line removed; why-shape bullet replaced.

Commit 2 — node.dag Practice-4 ledgers (T-1 closeout)

Gate (b) ratified: modeling-discipline.md §4 (coproduct dissolution) is a binding v4 rule, full ledger form. node.dag was the lone non-compliant outlier. Added the full 4-pattern 🟢 GREEN/terminal ledger to its 5 enums (Connective, Behavior, NodeKind, EdgeLabel, EdgeDiscipline) — comment-only; variants byte-unchanged (the closed kernel is terminal by C1 ratification; ledgers make that reviewable per the now-binding discipline).

Verification

• v2 bootstrap gate green at each commit: Part-A → 63 modules resolved (−5), 0 diagnostics; node.dag ledgers → 63 modules, 0 diagnostics.
• Diff vs main: 10 files, 136 ins / 143 del — exactly the intended change. No .dag substrate logic changed; node.dag variants byte-identical.

Test plan

• v2 gate 0-diag after deletions (nothing imported the cut files)
• v2 gate 0-diag after node.dag ledgers (comment-only)
• workflow/ = {bootstrap.dag, ci.dag}; bootstrap preserved (invariant-7 load-bearing)
• Authority-doc narrowing dated + honest (T-5 tombstoned, no silent retraction)
• Operator personal review of the THESIS facet-4 narrowing (parent-doc edit — flagged for your eyes specifically)

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 0a0f1fec · Trigger: manual
• Comparison: main @ a549f88c ... session/still-heron-594-cleanup @ 0a0f1fec
• Conversation: View conversation

---

1. Story of the diff

This PR narrows v4’s recursive-flex/meta-process claim from “model gunbc’s work-direction as .dag data” to “model only the build/bootstrap and CI pipeline as .dag data.” The thesis line is rewritten so meta-process modeling now names only workflow/bootstrap.dag and workflow/ci.dag, with the work-direction claim explicitly retracted as operator-ratified (THESIS.md:226, THESIS.md:260-280). The same semantic cut is propagated through worker authoring and structure docs: BRIEF_TEMPLATE.md removes the WorkerOutput DoD requirement and explains that the worker-output substrate was cut (src/v4/BRIEF_TEMPLATE.md:52-53, src/v4/BRIEF_TEMPLATE.md:104-110), STRUCTURE.md shrinks workflow/ to two files and updates the checksum from 78 to 73 files (src/v4/STRUCTURE.md:86-114), and TASKS.md tombstones T-5 while preserving bootstrap and CI as T-20/T-24 (src/v4/TASKS.md:181-198).

The code-modeling side of the PR is in src/v4/std/node.dag: it adds 🟢 terminal Practice-4 ledgers to the five closed substrate sum types: Connective, Behavior, NodeKind, EdgeLabel, and EdgeDiscipline (src/v4/std/node.dag:164, src/v4/std/node.dag:201, src/v4/std/node.dag:234, src/v4/std/node.dag:270, src/v4/std/node.dag:337). The five deleted workflow files were empty scaffolds for the now-retracted work-direction substrate, not implemented consumers (src/v4/workflow/brief.dag:14, src/v4/workflow/worker_output.dag:14).

2. Invariant categories

1. LAYER MODEL — Finding, BLOCKING.

The diff touches substrate declarations in src/v4/std/node.dag, so the terminal modeling evidence has to be complete. The newly added ledger for Connective says // 🟢 GREEN (terminal). Ledger — four patterns attempted: at src/v4/std/node.dag:164, and the same “four patterns” wording recurs for Behavior, NodeKind, EdgeLabel, and EdgeDiscipline at src/v4/std/node.dag:201, src/v4/std/node.dag:234, src/v4/std/node.dag:270, and src/v4/std/node.dag:337. Practice 4 requires trying five dissolution patterns before a terminal classification, including the parameterized-family/projection check; omitting that check makes the substrate GREEN classification under-evidenced. chatgpt-review-8ef26a4a-8175-4a…

1. INVARIANTS.md + modeling-discipline.md — Finding, BLOCKING.

Specific principle: P1 Modeling Faithfulness / Practice 4 Coproduct dissolution, with Practice 7’s projection-over-enumeration check as the missing fifth pattern. The diff’s ledgers enumerate only fact placement, variant-is-data, algebraic form, and dimensional checks; for example src/v4/std/node.dag:164-175 ends after “4. Dimensional” and immediately declares terminality at src/v4/std/node.dag:176-178. That skips the required “Parameterized family” attempt, which modeling-discipline calls out precisely because patterns 1–4 do not catch mechanically repeated F<X> families. chatgpt-review-8ef26a4a-8175-4a…

The fix is small but substrate-level: add a fifth bullet to each of the five ledgers, even if each says it fails trivially.

1. CODING.md — N/A.

N/A — this diff does not add or refactor Rust implementation code, helper placement, error/result shapes, methods, or hidden state; it changes docs, .dag comments/types, and deletes empty .dag scaffolds.

1. TESTING.md — N/A.

N/A — no executable behavior or test runner behavior changes here. The only .dag changes are ledger comments on existing substrate enums plus deletion of unfilled scaffold files, so there is no new behavior-driven test surface to add in this PR.

1. LOCKED DESIGN DECISIONS — Compliant.

The PR does alter a high-level thesis/design claim, but it does so explicitly rather than silently: the new thesis text says the facet was “narrowed 2026-05-15, operator-ratified” and then names the retracted work-direction sub-claim (THESIS.md:260-280). STRUCTURE.md likewise rewrites invariant 6 to say workflow/ is exactly bootstrap.dag and ci.dag, and explicitly retracts briefs / worker-outputs / cycles / retirement / doc-anchors as modeled process data (src/v4/STRUCTURE.md:293). That is the right shape for a locked-decision divergence: the meaning changes, but the change is surfaced at the authority layer.

1. TRACKED vs UNTRACKED DEBT — Compliant.

The PR removes scaffolds rather than adding new ones. The deleted files were explicitly marked Status: scaffold — fill per TASKS.md T-5 (src/v4/workflow/brief.dag:14, src/v4/workflow/cycle.dag:14, src/v4/workflow/doc_anchor.dag:13, src/v4/workflow/retirement.dag:14, src/v4/workflow/worker_output.dag:14), and TASKS.md now preserves the removal as a tombstone instead of silently dropping it (src/v4/TASKS.md:181-198). I do not see a new TODO, bridge, temporary carrier, or dissolution-trigger-less scaffold introduced by this diff.

2.5. Top-down PM intent review

Compliant, aside from the ledger completeness finding above. From the PM/executive-owner lens, the scope cut is not hidden semantic dilution: the PR explicitly says the project no longer models its own work-direction, preserves only bootstrap/CI pipeline self-application, and reconciles that same meaning across THESIS.md, STRUCTURE.md, TASKS.md, and BRIEF_TEMPLATE.md (THESIS.md:270-280, src/v4/STRUCTURE.md:86-91, src/v4/TASKS.md:181-198, src/v4/BRIEF_TEMPLATE.md:104-110). I do not see a diff-cited mismatch where a worker would faithfully execute the remaining plan and accidentally rebuild the retracted meta-layer; the plan now says not to.

The one PM-relevant risk is narrower: the PR title promises Practice-4 ledgers, but the added ledgers omit Practice 4’s fifth dissolution pattern. That is not a scope-cut disagreement; it is an incomplete modeling receipt on the remaining substrate work.

3. Verdict

REQUEST_CHANGES

The work-direction meta-layer cut is coherent, explicit, and consistently propagated. The blocker is that the new substrate GREEN ledgers claim only “four patterns attempted” even though the uploaded modeling discipline requires five; complete that fifth parameterized-family/projection check on all five node.dag ledgers before merging.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 0a0f1fec · Trigger: schedule
• Thinking: 239s wall

---

Non-blocking — Strengths

• src/v4/std/node.dag The Practice-4 ledgers classify the five substrate coproducts with concrete dissolution attempts and add no parallel authority.
• src/v4/TASKS.md The T-5 tombstone preserves the operator-ratified work-direction cut instead of silently dropping the deleted scaffolds.

✅ No blocking concerns; the narrowed build/CI workflow claim stays aligned with the thesis and v4 structure.

[briansrls]

Re: openai-pro/gpt-5.5-pro REQUEST_CHANGES (12713) — valid finding, fixed in e7f3baf31.

Verified against the current modeling-discipline.md on main: §4 was updated to require FIVE dissolution patterns (it added 5. Parameterized family + Practice 7 "projection over enumeration" — ~+4K chars; the codified lesson from the Homomorphism 15→1 dissolution). My 5 node.dag ledgers were written against the prior 4-pattern §4 and were under-evidenced. openai-pro is correct.

Added the 5th-pattern bullet to all five ledgers (Connective, Behavior, NodeKind, EdgeLabel, EdgeDiscipline). Each FAILS trivially with structural reasoning — the variants are heterogeneous (distinct payloads/arities/semantics), NOT one generic F<X> / projection over a source set (Practice 7) — so the 🟢 GREEN terminal classification stands, now fully evidenced per the updated rule. Comment-only; variants byte-unchanged. v2 bootstrap gate green: 63 modules resolved, 0 diagnostics.

Pre-existing follow-up flagged (NOT this PR's scope): the merged std/algebra.dag ledgers from #3154 (FreeMonoid, Ordering) also predate the 5-pattern §4 — they carry only 4 patterns and are now likewise under-evidenced against the updated rule. Since gate (b) is ratified binding-fully, that is a real consistency debt; it should be a small follow-up PR (add the 5th-pattern bullet to those two, same trivial-FAILS form). Out of scope here (this PR is the workflow-cut + node.dag), surfacing so it is tracked, not dropped.

— sent from still-heron-594

[briansrls]

openai-pro 12713 — now FULLY addressed (correcting my prior reply).

Re-verifying against current code (not assuming my earlier fix was complete) caught that e7f3baf31 added the 5th-pattern bullet to all 5 node.dag ledgers but left each lead-in saying Ledger — four patterns attempted: — internally inconsistent (announced four, listed five). That's exactly the under-evidenced shape openai-pro flagged, only half-fixed. Completed in c5b55018c: all 5 lead-ins → five patterns attempted:. So every node.dag ledger now: (a) lists patterns 1–5 incl. 5. Parameterized family (Practice 7 projection-over-enumeration, FAILS trivially with structural reasoning — variants are heterogeneous, not one generic F<X>), and (b) self-describes as five, fully compliant with the updated modeling-discipline.md §4. 🟢 terminal stands; comment-only; v2 gate green (63 modules, 0 diagnostics) at c5b55018c.

Verification receipt: grep -c "five patterns attempted" = 5, "four patterns attempted" = 0.

Pre-existing follow-up still tracked (NOT this PR): merged std/algebra.dag FreeMonoid/Ordering ledgers (from #3154) also predate the 5-pattern §4 — same trivial 5th-bullet + lead-in fix, separate PR.

— sent from still-heron-594