diff --git a/README.md b/README.md index db068d8..00b02fc 100644 --- a/README.md +++ b/README.md @@ -4,7 +4,7 @@ A framework for deciding **where determinations are supplied, how they are assured, and who answers for them** — at the level of individual decisions rather than whole jobs. -It applies to any arrangement that determines choices against ground: programs, models, humans, organisations, markets, immune systems, and compositions of them. +It is written to apply across arrangements that determine choices against ground — programs, models, humans, organisations, markets, immune systems, and compositions of them. \--- @@ -102,9 +102,9 @@ See [`core/01-the-principle.md`](core/01-the-principle.md) and [`core/02-complet \--- -## The central result: check the work or check the worker +## The central claim: check the work or check the worker -The framework's most important result concerns the **acceptance predicate**: the criterion by which an outcome would be judged acceptable. +The framework's central claim concerns the **acceptance predicate**: the criterion by which an outcome would be judged acceptable. The irreducible judgment floor is not a property of how "difficult" a task is. @@ -112,7 +112,9 @@ It is a property of whether adequacy can be checked. > **The floor is non-zero exactly when, and because, you cannot check the work.** -Where the acceptance predicate **closes for the arrangement** — relevant ground is observable and adequacy can be evaluated within the declared resource, latency, confidence, and assurance bounds — no particular determiner is required for trust. Different paths may produce acceptable outputs because the result itself can be checked. +*(Claim `DDD-floor-02`, **projected** — argued, with a declared falsifier that has not been met.)* + +Where the acceptance predicate **closes for the arrangement** — relevant ground is observable and adequacy can be evaluated within the declared resource, latency, and confidence bounds — no particular determiner is required for trust. Different paths may produce acceptable outputs because the result itself can be checked. Where the predicate does **not** close, that route to assurance disappears. @@ -120,7 +122,7 @@ Verification moves from the output toward the supplier: > **You cannot check the work, so you check the worker.** -This explains why organisations use licences, qualifications, grades, track records, certification, institutional standing, and selection. +This is the framework's reading of why organisations use licences, qualifications, grades, track records, certification, institutional standing, and selection — carried as a projected hypothesis (`DDD-hyp-05`), not a finding. They are not merely descriptions of capability. @@ -317,7 +319,7 @@ An open predicate within capacity remains judgment. Overflow on an open predicate produces escape. -This also yields a mechanical account of one class of hallucination: output decoupled from correct ground because that ground is absent, false, or present but unresolved. +Read as exposition, this suggests a mechanical reading of one class of hallucination: output decoupled from correct ground because that ground is absent, false, or present but unresolved. No claim carries that reading; the mechanism claim itself (`DDD-floor-01`) is `reported` on toy actors only. See [`core/11-the-floor-mechanism.md`](core/11-the-floor-mechanism.md). @@ -351,11 +353,11 @@ See [`core/07-determination-and-intelligence.md`](core/07-determination-and-inte \--- -## A worked instance outside software +## A suggestive parallel outside software -The actor-general vocabulary is tested against the vertebrate immune system. +The actor-general vocabulary is read against the vertebrate immune system. -The mapping is not intended as metaphor: +The reading is a **suggestive parallel with known disanalogies** — that demotion is on record, and no claim carries the instance at any status. The mapping offered: |Store|Immune-system instance| |-|-| @@ -364,11 +366,11 @@ The mapping is not intended as metaphor: |**Judgment**|adaptive per-encounter response| |**Escaped**|a pathogen no response catches| -The case also supplies an ensemble actor: diversity across a population carries coverage no single member can. +The case also suggests an ensemble actor: diversity across a population carries coverage no single member can. > **Redundancy buys reliability. Diversity buys coverage.** -The immune system matters because no engineer specified it. If the framework's categories only worked for deliberately engineered systems, the actor-general claim would fail here. +The parallel matters because no engineer specified the immune system: a vocabulary that could only describe deliberately engineered systems would not be actor-general. The instance is offered as a reading to be broken, not a test that was passed. See [`core/12-the-licensing-instance.md`](core/12-the-licensing-instance.md). diff --git a/core/05-accountability.md b/core/05-accountability.md index b900859..0ad032c 100644 --- a/core/05-accountability.md +++ b/core/05-accountability.md @@ -52,17 +52,29 @@ spectrum runs the other way. > **Accountability completeness is a property of an arrangement, not a capacity of an actor. An > execution is accountability-complete when it is linked to a persistent responsible principal, an -> attributable record, a defined stake, and an enforceable consequence path.** +> attributable record, a defined stake, an enforceable consequence path, and an authority with +> jurisdiction to enforce it.** > > It is independent of pinning resolution, in the sense that it is not recoverable from it — and it > is the scarcer of the two properties. +That is the completeness condition — five elements: attribution, persistent principal, authority +linkage, stake, sanction path *(claim `DDD-frame-08`, **projected**)*. It is not the definition. +The definition is the settled term embedded below, and the two are different objects. + > **Accountability** is a property of the arrangement, not of the executor: attribution of > the determination, a persistent answerable party, and a borne consequence. An arrangement > missing any of the three has not allocated the decision's consequence. +The definition and the completeness condition differ, and the difference has a test: an arrangement +can instantiate the relation and still be incomplete. Attribution exists, a persistent principal +exists, a consequence is borne — and no authority has jurisdiction to enforce it. Everyone knows who +is responsible; nobody can sanction them. The completeness condition refines the definition's borne +consequence into stake plus sanction path and adds authority linkage; the refinement and the +addition are carried at the claim's grade (**projected**), not the term's (**settled**). + A model does not simply *lack* sanctionability. Persistence depends on identity rules and continuity of obligation; stake depends on property, role, bond, reputation, or delegated interest; sanctionability depends on an authority with enforceable jurisdiction — all properties of a diff --git a/core/claims/DDD-frame-08.yaml b/core/claims/DDD-frame-08.yaml index b1f9622..3ebfbd9 100644 --- a/core/claims/DDD-frame-08.yaml +++ b/core/claims/DDD-frame-08.yaml @@ -2,13 +2,48 @@ format: 1 id: DDD-frame-08 kind: normative statement: | - Accountability is a relation (attribution, persistent principal, authority linkage, stake, sanction path), not an intrinsic capacity; an arrangement naming an executor but no principal is incomplete. + Accountability is a relation, not an intrinsic capacity of an executor. Completeness is a design-time condition on an arrangement, distinct from the relation's definition: a complete arrangement links attribution, a persistent principal, authority linkage, stake, and a sanction path. An arrangement can instantiate the relation and still be incomplete; an arrangement naming an executor but no principal is incomplete. status: projected region: consequential deployments test: >- stakeholder and case studies — completeness of the relation predicting trust and remediation outcomes better than executor identity owner: paper-4 -changed: v4.5 +changed: v5.14 (2026-09-03, pre-transfer session, GATE 1 ruling C — statement re-scoped as the completeness condition; status, region, test unchanged; version per the session's close proposal) notes: | - Conceptual core verified against canon: core/05 (v4.5, P1) states "Accountability completeness is a property of an arrangement, not a capacity of an actor... linked to a persistent responsible principal, an attributable record, a defined stake, and an enforceable consequence path", and core/01/core/05 §1 "a judgment allocation naming no accountable party is... Escaped with an executor attached". Kept projected: the normative prediction (relation-completeness predicting trust/remediation) is unexercised, needing the stated case studies. + Re-scoped, 2026-09-03 (pre-transfer session, GATE 1 ruling C, Emil). The prior statement read: + "Accountability is a relation (attribution, persistent principal, authority linkage, stake, + sanction path), not an intrinsic capacity; an arrangement naming an executor but no principal is + incomplete." Read against term:accountability (settled, three elements), that left two counts of + the relation's elements with nothing saying which governs. The ruling: the term defines what + accountability is — attribution of the determination, a persistent answerable party, a borne + consequence; this claim asserts what a complete instance requires at design time — the borne + consequence refined into stake plus the sanction path, and authority linkage added as a + completeness condition. The two objects differ, and there is a test: an arrangement can + instantiate the relation and still be incomplete — attribution exists, a persistent principal + exists, a consequence is borne, and no authority has jurisdiction to enforce it; everyone knows + who is responsible, nobody can sanction them. The term deliberately does not carry authority + linkage: a settled definition carries only what the corpus argues at settled grade, and authority + linkage's carriage is this projected claim and DDD-hyp-04's unrun prediction. core/05 §2 states + the distinction beside the term's embed. + + Conceptual core verified against canon, citation corrected in the same ruling: the prior notes + cited core/05 §2's completeness sentence alone, which carried four elements (persistent + responsible principal, attributable record, defined stake, enforceable consequence path) — a + four-element source cited as verification for five. The verification is the pair: that sentence + (now carrying all five, attributed to this claim), plus core/05 §2's sanctionability line — + "sanctionability depends on an authority with enforceable jurisdiction — all properties of a + relationship among executor, principal, record, and sanctioning authority" — which is where the + prose argues the fifth element. Also core/01/core/05 §1: "a judgment allocation naming no + accountable party is... Escaped with an executor attached". + + Lineage (Paper A's related-work survey, 2026-08-30): four of the five elements are already + Bovens's — the actor–forum relation, with the actor obliged to inform, the forum able to + interrogate, and the forum able to pass judgement and impose sanctions (European Law Journal + 13(4), 2007). The delta is authority linkage, which Bovens does not carry as a distinct element, + and a change of tense: Bovens's relation is assessed retrospectively over conduct, this claim's + completeness is constituted at design time. Half the delta is a tense change and is claimed as no + more than that. + + Kept projected: the normative prediction (relation-completeness predicting trust/remediation) is + unexercised, needing the stated case studies. diff --git a/core/claims/DDD-measure-11.yaml b/core/claims/DDD-measure-11.yaml index 03fb7f3..4918c74 100644 --- a/core/claims/DDD-measure-11.yaml +++ b/core/claims/DDD-measure-11.yaml @@ -6,7 +6,7 @@ statement: > function and the deployment distribution and says nothing about the cost of computing a correct answer. Two tasks with identical verdict entropy can differ unboundedly in generation cost, so the measure must not be read as pricing generation. -status: reported +status: projected region: closing predicates only (where H(verdict) exists) credits: > Cook-Levin (1971), via DDD-frame-06: satisfiability is linear to check yet NP-hard to @@ -21,9 +21,15 @@ falsifier: > A task family in which verdict entropy bounds generation cost, so that the measure does price the search. owner: paper-1 -changed: v4.5 +changed: v5.14 (2026-09-07, pre-transfer session, GATE 4 ruling — demoted reported -> projected; statement, region, evidence and falsifier unchanged; version per the session's close proposal) notes: > Extracted in the core/09 conversion; added to canon at v4.5 (generation-cost correction). The measure-framed companion of DDD-frame-06 (the closure/generation-cost separation, established) — the two variables this release separates must not be re-fused through the measure. + + Demoted reported -> projected, 2026-09-07 (pre-transfer session, GATE 4 ruling, Emil): + the only evidence entry is a derivation, and spec/claim-format.md §5 defines reported as + carrying at least one evidence entry whose asset reproduces. The derivation stays + recorded; the status now matches what the evidence carries — and matches how the primer's + §6 already read this claim. Restatus to reported when a reproducing asset exercises it. diff --git a/core/claims/DDD-measure-13.yaml b/core/claims/DDD-measure-13.yaml index 570b5dc..62171c7 100644 --- a/core/claims/DDD-measure-13.yaml +++ b/core/claims/DDD-measure-13.yaml @@ -6,7 +6,7 @@ statement: > decomposition harvests more of the verdict, runtime demand falls toward the residual the encoding has not captured, never below what the open part of the predicate leaves undetermined. -status: reported +status: projected region: closing predicates only, conditional on DDD-measure-01 credits: The theorem is Shannon's (chain rule, 1948); only the identification is ours. evidence: @@ -15,8 +15,14 @@ evidence: note: §5, third bullet — the asymptote of the maturation/funnel projection (core/08) in measure terms falsifier: none — arithmetic given the chain rule; fires with DDD-measure-01 owner: paper-1 -changed: v4.5 +changed: v5.14 (2026-09-07, pre-transfer session, GATE 4 ruling — demoted reported -> projected; statement, region, evidence and falsifier unchanged; version per the session's close proposal) notes: > Extracted in the core/09 conversion (§5). Connects the measure to the maturation/funnel projection in core/08-projections.md; the seed landed neither the asymptote formula nor that link. + + Demoted reported -> projected, 2026-09-07 (pre-transfer session, GATE 4 ruling, Emil): + the only evidence entry is a derivation, and spec/claim-format.md §5 defines reported as + carrying at least one evidence entry whose asset reproduces. The derivation stays + recorded; the status now matches what the evidence carries — and matches how the primer's + §6 already read this claim. Restatus to reported when a reproducing asset exercises it. diff --git a/releases/v5.14.0.yaml b/releases/v5.14.0.yaml new file mode 100644 index 0000000..0ac26c3 --- /dev/null +++ b/releases/v5.14.0.yaml @@ -0,0 +1,70 @@ +format: 1 +version: v5.14.0 +# No `commit`: cut wherever this file lands on the default branch, the normal case +# (spec/release-format.md §2 rule 3). +date: 2026-09-07 +basis: + # The claim re-scoped as the completeness condition (GATE 1 ruling C): the settled term + # defines the relation; the claim asserts what a complete instance requires at design time. + - DDD-frame-08 + # Examined under the same ruling and deliberately untouched: its prediction already + # quantifies over completeness of the five-element arrangement. + - DDD-hyp-04 + # The two demotions: derivation-only evidence under a status §5 defines as requiring a + # reproducing asset (GATE 4 ruling). + - DDD-measure-11 + - DDD-measure-13 +title: 'Publication readiness — the arity ruled, the front page carries no more than a status does' +summary: | + **The last canon work before the repositories become public**, chosen by one test: does this get + more expensive, or become unfixable, once someone outside can read it or fork it? Interactive + throughout; every change below is a gate ruling (pre-transfer session, Emil; the session record + is the projection repository's `meta/sessions/2026-09-02-pre-transfer/`). + + This is a minor and not a patch: a projected claim's statement moves by supersession and two + statuses move. It is not a major because every format-1 claim remains valid unchanged, no id + moves, and no settled node moves. + + ### The accountability arity, ruled (disposition C) + + `DDD-frame-08` (projected) counted five elements where `term:accountability` (settled) counts + three, and nothing said which governs. Ruled: **they are different objects, and the difference + has a test.** The term defines what accountability is — attribution, a persistent answerable + party, a borne consequence. The claim asserts what a *complete* instance requires at design + time — the borne consequence refined into stake plus sanction path, and authority linkage added + as a completeness condition. An arrangement can instantiate the relation and still be + incomplete: everyone knows who is responsible; nobody can sanction them. The claim's statement + now says so; the term is untouched; `core/05` §2 states the distinction beside the embed, with + its completeness sentence carrying all five elements at the claim's grade, attributed. The + lineage lands in the claim's notes at its true size: four of the five elements are Bovens's; + the delta is authority linkage plus the change from retrospective assessment to design-time + constitution — half of it a tense change, claimed as no more. + + ### Two statuses stop overstating their evidence + + `DDD-measure-11` and `DDD-measure-13` demote `reported` → `projected`: their only evidence is a + derivation, and `spec/claim-format.md` §5 defines `reported` as carrying at least one evidence + entry whose asset reproduces. The derivations stay recorded; the downstream primer already read + both conservatively as projected, and canon now agrees with its own projection. Neither is + pinned downstream; nothing fires on their account. + + ### The front page carries no more than a status does + + Repaired in place, converging on the register the README already used where it was right: the + central claim is named a claim with `DDD-floor-02`'s status beside it; the licences sentence + carries `DDD-hyp-05` as a projected hypothesis, not a finding; the hallucination passage is + marked exposition; the immune-system section takes the recorded v3-era demotion — a suggestive + parallel with known disanalogies, offered as a reading to be broken, not a test that was + passed; the applicability sentence states intent rather than asserting universality; and the + closure sentence returns to the settled term's three bounds — the fourth was the front page's + own addition, which three sources contradicted. + + ### What fires, stated before it happens + + When the downstream pin next advances past this release, **exactly one W6 fires**: + `DDD-frame-08`, `sha256:cf73b307…` → `sha256:ef5d5faf…` (statement re-scoped, status held). + The demotions fire nothing — `DDD-measure-11` and `DDD-measure-13` are not pinned, and status + is instrumented by W5 only for pinned ids. Zero W5, zero W7 change. The downstream pin advance + *to* v5.13.0 was executed and verified in the same session (downstream `DDD-dec-35`): the + ground migration's own two-W6 prediction held in every limb, by id and by hash, across a tag + and a merge. diff --git a/spec/claim-format.md b/spec/claim-format.md index fe0a41f..050e4f5 100644 --- a/spec/claim-format.md +++ b/spec/claim-format.md @@ -1,7 +1,8 @@ # Claim format specification **Format version: 1.** This file is the versioned schema artifact for claim nodes — shape and -validation rules only. It contains no process (see `meta/conversion-protocol.md`, versioned with +validation rules only. It contains no process (see the projection repository's +`meta/conversion-protocol.md`, versioned with the way of working) and no content (see `core/claims/`, versioned per claim via `changed`). The three change independently: