Skip to content

Commit 52d5b03

Browse files
docs(machine-readable): update STATE + META for F1/F3/earn-back closure (#93)
## Summary Brings `.machine_readable/6a2/STATE.a2ml` and `META.a2ml` in line with the prose docs after this session's Pillar F earn-back closure and §"Theory work" closing. ## `META.a2ml` — three new ADRs (append-only) No existing ADR is edited. Three appended after `adr-006`: * **`adr-007`** — F1 earn-back via monoid-graded iterated-residue construction. `EchoGradedComonadF1.agda` at `(ℕ, +, 0)` with Echo as grade-unit object, nested δ, all three comonad laws, `D2-nontrivial` separating witness. (PR #86 merged 2026-05-20.) * **`adr-008`** — F3 earn-back via two non-isomorphic-grade-monoid instances of an abstract `GradedComonadStructure` interface. No `⊑-prop`-equivalent field. `nat-instance` at commutative `(ℕ, +, 0)`; `list-instance` at non-commutative `(List Tag, ++, [])`. (PR #88 open.) * **`adr-009`** — Retraction-discipline succeeded: the R-2026-05-18 reframing converted into four earn-back gate passes (F4, F2, F1, F3); one retraction stays retracted (conservativity metatheorem, finding 5); none silently re-inflated. Each ADR carries the strict scope qualifier — F1/F3 earn back claims **for a separate side-construction**; `EchoGraded` itself remains a thin-poset reindexing modality per R-2026-05-18, and `paper.adoc` / `types-abstract.adoc` / `conservativity.adoc` are intentionally not moved. ## `STATE.a2ml` — `next-actions` pruned + two closure sections appended **Pruned**: 3 of 5 April `next-actions` were stale and are resolved in the prose: * `integration` (apply 7-commit sequence) — done long ago * `t-3` (Gate-1 falsification test 3) — superseded by Gate 1 adjacency refresh (`decisions/gate1-adjacency-refresh.adoc`, PR #77) * `q2-4` (falsifier attempts) — absorbed by `IntegrationAudit.agda` EI-2 negative result; remaining attempts are bookkeeping **Remaining + added**: `q2-1` (echo-not-prop generalisation, still high), `q2-3` (RoleGraded as N5, still low), new `owner-gated-paper-update` entry, new `ordinal-track-path-1` entry (in-flight in other session), and the original `v0-2-recipe-extension` parked entry. **Two new closure sections** at the end of STATE.a2ml: * **`earn-back-summary`** — closure record for all four gates with module paths, claim wording, retraction-followup citations, and a `forbidden-rebrandings` list to prevent later mis-framing. * **`theory-work-summary`** — closure record for §"Theory work — no proof assistant needed": axes fully mechanised list, ruled-out items, refreshed items, canonical examples cluster. ## Test plan * [x] Pure machine-readable data; no Agda touched * [x] `agda proofs/agda/All.agda` exit 0 * [x] `agda proofs/agda/Smoke.agda` exit 0 * [x] Append-only ADR discipline preserved (adr-001 through adr-006 unchanged; new ADRs at the end) * [x] No retracted claim moved; no earned-back claim overclaimed * [ ] (owner) confirm a2ml schema validation if you run a validator against the standards/state-a2ml + meta-a2ml specs ## Scope Pure machine-readable data update. Disjoint from all currently-open PRs (#88 F3, #90 partition-dec, #91 generic-codomain, #92 docs tidy). `paper.adoc` / `types-abstract.adoc` / `conservativity.adoc` are intentionally not touched. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
1 parent 0fb7587 commit 52d5b03

2 files changed

Lines changed: 126 additions & 18 deletions

File tree

.machine_readable/6a2/META.a2ml

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -71,6 +71,33 @@
7171
(context . "User works across two machines: Fedora Kinoite (primary, Nushell) and a Windows machine for travel.")
7272
(decision . "Build instructions are path-agnostic; line endings normalised via .gitattributes (LF for Agda/AsciiDoc/YAML, .agdai marked binary). Cross-platform considerations apply throughout.")
7373
(consequences . "Path-agnostic build instructions are required, not optional, for any new tooling.")
74+
(supersedes . none))
75+
76+
(adr-007
77+
(title . "F1 earn-back via monoid-graded iterated-residue construction")
78+
(status . accepted)
79+
(date . "2026-05-20")
80+
(context . "R-2026-05-18 retracted the graded-comonad claim about EchoGraded — the structure was a thin-poset reindexing with no nested family, no monoid multiplication, no genuine δ. Pillar F gate F1 (docs/echo-types/earn-back-plan.adoc §F1) asked whether ANY genuine graded comonad with Echo as the grade-unit object could be mechanised under --safe --without-K with zero postulates.")
81+
(decision . "The candidate construction passes: proofs/agda/EchoGradedComonadF1.agda ships a monoid-graded iterated-residue comonad at the grade monoid (ℕ, +, 0) with D 0 A = A; D (suc r) A = R (D r A) where R X = X × Bool (an informative residue layer, not ⊤). All three graded-comonad laws (gc-counit-l, gc-counit-r, gc-coassoc) typecheck under --safe --without-K with zero postulates; gc-coassoc closes via the predicted δ-naturality-over-R factoring (δ-suc + subst-D-suc). The separating witness D2-nontrivial certifies D r is not collapsing to ⊤ / a prop.")
82+
(consequences . "F1 PASSED; the graded-comonad claim is earned back FOR THIS WITNESS ONLY. EchoGraded itself remains a thin-poset reindexing modality per R-2026-05-18 — F1 enters as an *additional* mechanised contribution beside EchoGraded, not as a reinstatement of it. Paper title and central thesis (Echo as a reindexing modality) stand unchanged. Unblocks F3 (independent second comonad model). Retraction follow-up F-2026-05-20a appended to docs/retractions.adoc.")
83+
(supersedes . none))
84+
85+
(adr-008
86+
(title . "F3 earn-back via two non-isomorphic-grade-monoid instances of an abstract interface")
87+
(status . accepted)
88+
(date . "2026-05-20")
89+
(context . "F1 (adr-007) earned back the existence of a graded comonad with Echo as grade-unit object. Gate F3 (docs/echo-types/earn-back-plan.adoc §F3) asked whether the construction is genuinely model-independent — instantiable at non-isomorphic grade monoids without a single hypothesis (no ⊑-prop-equivalent field) baking in the result.")
90+
(decision . "EchoGradedComonadInterface.GradedComonadStructure is an abstract record packaging the F1 signature (grade monoid + monoid laws + graded functor + functor laws + counit + nested δ + the three comonad laws stated against subst along the monoid's propositional identities). The record carries NO ⊑-prop-equivalent field — only structure, monoid laws, and comonad laws. Two non-isomorphic-grade-monoid instances inhabit it: EchoGradedComonadInstance1.nat-instance at the commutative monoid (ℕ, +, 0); EchoGradedComonadInstance2.list-instance at the non-commutative free monoid (List Tag, ++, []) over a two-element Tag with per-element residue layers R smol A = A × Bool and R big A = A × ℕ. Non-isomorphism is constructively witnessed by tag-list-non-commutative.")
91+
(consequences . "F3 PASSED; the two-models claim is earned back FOR THE F1-STYLE GRADED-COMONAD WITNESS. It does NOT reinstate the older EchoRelModel/GCLaws two-models claim retracted at R-2026-05-18 finding 3 — that situation (same grade poset, ⊑-prop baked in as a field, rel-model = set-model × ⊤, agreement by refl) is unchanged. The two earn-backs are about different abstract interfaces and are not interconvertible. Retraction follow-up F-2026-05-20b appended to docs/retractions.adoc. Pillar F earn-back programme now CLOSED: F4 + F2 (2026-05-18), F1 + F3 (2026-05-20).")
92+
(supersedes . none))
93+
94+
(adr-009
95+
(title . "Retraction-discipline succeeded: R-2026-05-18 reframing converted into four earn-back gate passes")
96+
(status . accepted)
97+
(date . "2026-05-20")
98+
(context . "R-2026-05-18 retracted five claims and reframed the project around what the Agda actually shows (thin-poset reindexing modality, not graded comonad; pointwise mediator, not terminal cone; carrier-parametricity, not model-independence; postulate-free build as evidence, not conservativity metatheorem; no funext anywhere, not 'quarantined'). The earn-back plan in docs/echo-types/earn-back-plan.adoc was the falsifiable program for converting the retracted claims back into theorems — or confirming, on the project's own gate discipline, that they cannot be earned at their original strength.")
99+
(decision . "All four gates have now passed at the strictly-bounded strength the earn-back plan asked for: F4 (terminal-cone UP as a function of an explicit funext parameter, never a postulate); F2 (genuine second model of the bare Echo functor on a non-graph StepND relation); F1 (genuine graded comonad on iterated-residue carrier with Echo as grade-unit object); F3 (the F1 construction is instantiable at non-isomorphic grade monoids). Each is exactly as strong as the gate specified; nothing is overclaimed. The conservativity metatheorem retraction (R-2026-05-18 finding 5) stays retracted with no gate attempting to earn it back — that one was a meta-statement over all propositions, not discharged by typechecking. The 'not two models for EchoGraded/GCLaws' finding 3 also stays retracted; the F3 earn-back is about the different F1 interface.")
100+
(consequences . "The retraction discipline is validated AS A METHODOLOGY: a retraction is not a failure but the mechanism by which claims become falsifiable. Four of five retracted claims were earned back at honest strength, one stays retracted, none was silently re-inflated. paper.adoc / types-abstract.adoc / conservativity.adoc are NOT moved by this ADR — those documents are about EchoGraded's thin-poset structure, which the F1+F3 side-construction does not change. Whether to add a bounded 'new contribution' paragraph is owner-gated.")
74101
(supersedes . none))))
75102

76103
;;; ============================================================

.machine_readable/6a2/STATE.a2ml

Lines changed: 99 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -234,19 +234,20 @@
234234
;;; ============================================================
235235

236236
(define next-actions
237-
'(((id . t-3)
238-
(title . "Gate-1 falsification test 3 (informativeness collapse)")
239-
(status . unstubbed)
240-
(relevance . "Now better-targeted post-EI-2: with RoleGraded in place, ask whether the role-or-grade axis is derivable from the other plus a forgetful functor.")
241-
(priority . medium))
242-
243-
((id . q2-4)
244-
(title . "Falsifier attempts for surviving gate-2 nominees")
245-
(status . open)
246-
(relevance . "Each surviving nominee has an unattempted falsifier. Worth attempting at least one to harden the gate-2 case beyond 'no successful reduction has been attempted'.")
247-
(priority . medium))
248-
249-
((id . q2-1)
237+
;; Updated 2026-05-20: the April 2026 next-actions list is largely
238+
;; superseded by intervening session activity. Resolutions:
239+
;; * `integration` (apply 7-commit integration sequence) — DONE,
240+
;; long since merged.
241+
;; * `t-3` (Gate-1 falsification test 3) — superseded by the
242+
;; Gate 1 adjacency refresh (`decisions/gate1-adjacency-refresh.adoc`,
243+
;; PR #77) which closed gate-1 work at 5/5 REFINED.
244+
;; * `q2-4` (falsifier attempts) — partially absorbed into
245+
;; `IntegrationAudit.agda` EI-2 negative result; further
246+
;; falsifier attempts are now bookkeeping not load-bearing.
247+
;; * `q2-1` (echo-not-prop generalisation) — still open.
248+
;; * `q2-3` (RoleGraded as N5) — still open (low priority).
249+
;; * `v0-2-recipe-extension` — still parked.
250+
'(((id . q2-1)
250251
(title . "Generalisation of echo-not-prop")
251252
(status . open)
252253
(relevance . "n=2 special case proofs generalise to: for non-injective f with at least two distinct preimages of y, is-prop (Echo f y) → ⊥. Closing this also discharges the truncation-argument gap for refinement, IFC, and provenance adjacency notes.")
@@ -260,18 +261,98 @@
260261
(priority . low)
261262
(note . "Mostly bookkeeping post-EI-2 since the recipe is no longer the distinctness locus."))
262263

263-
((id . integration)
264-
(title . "Apply the seven-commit integration sequence")
265-
(status . ready)
266-
(relevance . "INTEGRATION_COMMITS.adoc has the full sequence. Commit 7 is the EI-2 termination. All artefacts are staged at /mnt/user-data/outputs/audit-output/.")
267-
(priority . high))
264+
((id . owner-gated-paper-update)
265+
(title . "Bounded new-contribution paragraph in paper.adoc / types-abstract.adoc / conservativity.adoc for F1+F3")
266+
(status . owner-gated)
267+
(relevance . "F1 + F3 (see META adr-007, adr-008) introduce a separate graded-comonad construction beside EchoGraded. The paper bodies are about EchoGraded's thin-poset structure — the title and central thesis should not move on these gates alone. Whether to add a bounded contribution paragraph mentioning the F1/F3 side-construction is an editorial decision.")
268+
(priority . deferred))
269+
270+
((id . ordinal-track-path-1)
271+
(title . "Ordinal track Path-1 — Brouwer rank-mono into _<ᵇ⁻_")
272+
(status . in-flight)
273+
(relevance . "Brouwer arithmetic + rank-mono for _<ᵇ⁻_ are in a parallel-session worktree; this echo-types session stays clear of `proofs/agda/Ordinal/**` per `feedback_parallel_session_branch_drift` discipline.")
274+
(priority . high)
275+
(owner . other-session))
268276

269277
((id . v0-2-recipe-extension)
270278
(title . "Recipe extension exploration (parked)")
271279
(status . parked-v0.2)
272280
(relevance . "Coupled state across axes; multiple live positions per axis. New module work, not a re-investigation of EI-2.")
273281
(priority . deferred))))
274282

283+
;;; ============================================================
284+
;;; Pillar F earn-back closure summary (2026-05-20)
285+
;;; ============================================================
286+
;;;
287+
;;; The Pillar F earn-back programme launched by R-2026-05-18 is now
288+
;;; CLOSED. All four gates passed at strictly-bounded strength; see
289+
;;; META.adr-{007,008,009} and docs/retractions.adoc follow-ups
290+
;;; F-2026-05-18a + F-2026-05-20a + F-2026-05-20b.
291+
292+
(define earn-back-summary
293+
'((status . closed)
294+
(closing-date . "2026-05-20")
295+
(gates
296+
((F4 (status . passed) (date . "2026-05-18")
297+
(module . "proofs/agda/EchoPullbackUnivF4.agda")
298+
(claim . "Terminal-cone universal property as a function of an explicit funext module parameter, never a postulate")
299+
(retraction-followup . "F-2026-05-18a"))
300+
(F2 (status . passed) (date . "2026-05-18")
301+
(module . "proofs/agda/EchoStepNDModelF2.agda")
302+
(claim . "Genuine second model of the bare Echo functor on the non-graph relation StepND; content-bearing agreement")
303+
(retraction-followup . "F-2026-05-18a"))
304+
(F1 (status . passed) (date . "2026-05-20")
305+
(module . "proofs/agda/EchoGradedComonadF1.agda")
306+
(claim . "Monoid-graded iterated-residue comonad at (ℕ, +, 0) with Echo as grade-unit object; nested δ; all three comonad laws; D2-nontrivial separating witness")
307+
(retraction-followup . "F-2026-05-20a")
308+
(adr . adr-007))
309+
(F3 (status . passed) (date . "2026-05-20")
310+
(modules . ("proofs/agda/EchoGradedComonadInterface.agda"
311+
"proofs/agda/EchoGradedComonadInstance1.agda"
312+
"proofs/agda/EchoGradedComonadInstance2.agda"))
313+
(claim . "Abstract GradedComonadStructure record (no ⊑-prop-equivalent field) + two non-isomorphic-grade-monoid instances: nat-instance at commutative (ℕ, +, 0), list-instance at non-commutative free monoid (List Tag, ++, [])")
314+
(retraction-followup . "F-2026-05-20b")
315+
(adr . adr-008))))
316+
(scope-qualifier . "F1+F3 earn back the graded-comonad and two-models claims FOR A NEW SIDE-CONSTRUCTION. EchoGraded remains a thin-poset reindexing modality per R-2026-05-18; EchoRelModel's two-models claim remains retracted. Different abstract interfaces; not interconvertible.")
317+
(forbidden-rebrandings
318+
. ("EchoGraded is a graded comonad"
319+
"the F1 construction reinstates EchoGraded's retracted comonad claim"
320+
"the F3 two-models result reinstates EchoRelModel's retracted two-models claim"
321+
"the title or central thesis of paper.adoc moves on F1 or F3 alone"))
322+
(unmoved-retractions
323+
. ("R-2026-05-18 finding 5 (conservativity metatheorem) — stays retracted; no gate attempted to earn it back, no gate could (meta-statement over all propositions, not discharged by typechecking)"
324+
"R-2026-05-18 finding 3 (\"Not two models\" for EchoRelModel/GCLaws) — stays retracted; F3 earned a different two-models claim at a different interface"))))
325+
326+
;;; ============================================================
327+
;;; Theory-work closure summary (2026-05-20)
328+
;;; ============================================================
329+
;;;
330+
;;; The §"Theory work — no proof assistant needed" section in
331+
;;; docs/echo-types/roadmap.md is now empty. Every item is landed,
332+
;;; ruled out, or refreshed.
333+
334+
(define theory-work-summary
335+
'((status . closed)
336+
(closing-date . "2026-05-20")
337+
(closure-prs . (#67 #68 #69 #70 #71 #72 #74 #75 #76 #77 #78 #79 #82 #84 #86 #88 #90 #91 #92))
338+
(axes-fully-mechanised
339+
. ((axis-2-approximate-echo . "EchoApprox.agda + EchoApproxInstance.agda; Rung C BalancedTolerance at #78")
340+
(axis-8-decidability . "EchoDecidable.agda")
341+
(axis-8-cost . "EchoCost.agda + EchoCostInstance.agda (PR #85)")
342+
(axis-8-access . "EchoAccess.agda (PRs #68, #75)")
343+
(axis-8-search . "EchoSearch.agda + EchoSearchInstance.agda (PR #80)")
344+
(negative-coecho . "AntiEcho.agda (PR #69) + AntiEchoTropical.agda (PR #72) + AntiEchoTropicalGeneric.agda (PR #91) + antiecho-partition-dec (PR #90)")))
345+
(ruled-out
346+
. ((two-categorical-shape . "decisions/no-2-cat.adoc — every would-be 2-cell is refl or forced trivial")))
347+
(refreshed
348+
. ((presentation-dependence-cluster . "decisions/presentation-dependence.adoc — examples 5, 9, 10 cluster; meta-pattern only, no new module needed")
349+
(gate-1-adjacency-refresh . "decisions/gate1-adjacency-refresh.adoc — 5/5 REFINED, 0 RE-EVALUATE")))
350+
(canonical-examples-cluster
351+
. ((example-5-database-provenance . "EchoExampleProvenance.agda (PR #81)")
352+
(example-9-parser . "EchoExampleParser.agda (PR #83)")
353+
(example-10-abstract-interp . "EchoExampleAbsInt.agda (PR #82)")
354+
(example-6-numeric . "still unblocked, only remaining example item")))))
355+
275356
;;; ============================================================
276357
;;; END OF STATE
277358
;;; ============================================================

0 commit comments

Comments
 (0)