inputs-gpt-execution-adjudication-v5.md)package-v2.8/ supersedes v2.7 (structure ACCEPTED; evidence state patched here) — an evidence-state patch plus four small executions (v5 §6, §15). diff -r package-v2.7 package-v2.8 is the delta; derived by tools-lane/build_v2_8.py (every anchor exactly once — audited against the v2.7 documents before the run; every receipt hash recomputed; every expectation count, series result and mutation count derived by running or by reading the shipped artifact) plus tools-lane/drift_gate_v2_8.py. The v2.8 gate PASSES on v2.8, FAILS on v2.7 (63 findings) and FAILS under forty-one in-memory mutations (tools-lane/gate-v2_8-negative-control.txt, refusals counted by exit code). The gate also runs the station receipt's verify.
| what | where | result | ||
|---|---|---|---|---|
| R0.8.8 — requirement e as NC31, in the computedriven repo (this session acting as that lane) | px13/COMPUTEDRIVEN_R088_FOR_GPT.{md,tar.gz} (5.9 MB, sha256 0253b23c…; commit e42cc88, tree 918b7ee0…) | the contract stated first (docs/durable-authority-model.md), then the control: not clean at `89adeb3` — F38, the trusted report said clean while recovery cancelled the unused reservation (after‑anchor‑pending) and while it advanced the lineage (a NEW after-root-finalize fault point, so the F35 window is reached by a real crash, not by construction); 16 "did it say so" gates failing while every "did it happen" gate passed; repaired at e42cc88 (AnchorRecovery, total over the arms that write; three in-crate controls incl. read-only); ten legs, 21/21, verify=0 dirty=0; falsification committed before repair | ||
| E-51 → v2.8 + E-52 effect-time revalidation (v5 §11) | experiments/s5_effect_adapter/effect_identity_model.py (v2.7 kept as .v2_7.py / .v2_7.output.txt) | 27/27, seed-stable: the AVOIDABLE duplicate is unreachable under J1 for every far side and reconciler; the acknowledged-risk duplicate stays reachable; the v2.7 claim-time-only rule kept as a mutation that reintroduces it; revalidation is a second line, not a substitute for the claim rule; revalidation needs a far side that can say satisfied | ||
| EXP-7 v6 — the outside compare-and-mutate as a TOCTOU (v5 §13) | experiments/trvm_contracts/probe_contracts_e42_e43.mjs (v5 output kept as probe_contracts.v5.output.txt) | 40/40: outside → two transitions for one change and a stale no-op; inside (one World operation) → one transition, the stale no-op impossible, an identical re-registration under the maintenance lock refused by name | ||
the factory kit v6 — evidence-state@2, the station receipt | tools-lane/factory-kit/ | validator selftest 21 mutations / 37 applications over the @2 and @1 documents, all refused; `station-receipt.mjs make | verify | --selftest` (every single-binding mutation refused); both claim records updated; the v2.8 package is the first composition |
| record | v2.7 | v2.8 | adjudication |
|---|---|---|---|
| E-13b | CLOSED-F35; F36/F37 repaired-unadjudicated | + CLOSED-F36 / CLOSED-F37; cross_references → S2 (recovery may not mint authority from claimant state); corollary_name S3-PHASE-COMPLETE-PROVENANCE (a refinement, not a primitive); a closure transition (token unchanged) typed with the R0.8.7 receipt, subject_identity 21a1452, adjudication_ref v5 §1–§3; F38 stated as repaired-unadjudicated | v5 §1–§3 |
r0_8 | F36/F37 open; freeze_decision PENDING | closed_by_adjudication_v5 (F36, F37); f36_status/f37_status CLOSED; `nc31_status` EXECUTED (R0.8.8, F38 repaired, UNADJUDICATED); f38_status; requirements map item e EXECUTED; `safety_mechanism_status` READY FOR FREEZE; `freeze_decision` PROPOSED within the explicit profile — not asserted; the S2 cross-reference and the corollary name recorded | v5 §1–§5 |
| E-51 | 19/19 | re-executed v2.8 (27/27) — an evidence_refresh transition (token unchanged); the v2.7 receipt kept append-only | v5 §11 |
| E-52 (new) | — | S5 / corollary / EXHAUSTIVE-IN-MODEL: effect-time currentness / revalidation before the irreversible boundary; a class_promotion transition whose witness is the v2.7 rule kept as a mutation; minted from the executed model, not from prose | v5 §11–§12 |
| E-53 (new) | — | S4 / corollary / FALSIFIED at 89adeb3 (F38), repaired at e42cc88, repaired_unadjudicated; pre-fix + repair witnesses; no transition until adjudicated — the gate refuses one | v5 §4 |
| E-49 | 35/35 | 40/40 (the TOCTOU witness); v5 §13 recorded | v5 §13 |
| L1? | three assumptions | + discharge_structure_v5: SYS-SUCCESSOR-ENABLEMENT stays ONE obligation, DISCHARGED_BY authority evidence / policy admission / finality-generation proof / resource compatibility; split only on an independence witness | v5 §10 |
promotions[] | 5 entries | `evidence_transitions[]` (6 entries), each with a transition_kind that must agree with its tokens: 3 class_promotion (E-15, E-48, E-52) · 2 closure (E-13b, E-14) · 1 evidence_refresh (E-51); promotions_renamed_v2_8 recorded; the projection is evidence-state@2 | v5 §8 |
station_receipt (new) | — | station_receipt.json made by the kit: input registry (canonical, without the pointer) → generator → transitions → projection → validator → result → subject; the registry points at the receipt; the gate verifies it | v5 §7 |
factory_candidates | SCHEMA-CLOSED drafted | FAC-EVIDENCE-TRANSITION-SCHEMA-CLOSED PROVED / exhaustive-in-declared-domain (evidence-state@1 ∪ @2; counts derived — the v2.7 claim text hand-typed a count one higher than its list); FAC-CONTROL-SENSITIVITY DECLARED + the station-receipt rule | v5 §8–§9 |
cross_cutting_observation_v2_8 (new) | — | admission at proposal time ≠ currentness at use time — F35 / F36 / F37 / F38 / E-52 / EXP-7 v6 tabulated (frontier §9, synthesis §9); OBSERVED, not minted; the gate refuses a promotion of it | for adjudication |
frontier: header; S3 / S4 / S5 rows; §5 items 7–8; §7 (the discharge structure); §8 (PROVED narrow claim, the rename, the receipt); new §9; synthesis / experiments / handoff (B, C, E, F, G) / README correspondingly; EXTERNAL_RESEARCH_PINS_V4.md (ARIES compensation log records; leases; SLSA / in-toto provenance; the IETF Idempotency-Key draft; check/use races) beside the v3 pins.
handoffs/SUPER_E50_EFFECT_IDENTITY.md v6 addendum (E-52 in ampd's vocabulary); handoffs/TRVM_E42_E43_CANDIDATE_LAWS.md v6 addendum (the TOCTOU table; decision drafts A-IDEMPOTENT-GUARDED inside the World and fork() = descendant with recorded ancestry, each with a native-witness spec); tools-lane/factory-kit/README.md v6.
px13/ ships `COMPUTEDRIVEN_R088_FOR_GPT.{md,tar.gz}` whole; px13/README.md carries the v6 paragraph — including the provenance note that this session acted as both lanes this round; SHIPPED-SHA256SUMS.txt regenerated last.
Receipt-bytes convention, stated: the E-51 witness path is the current effect_identity_model.output.txt; v2.7's E-51 witness hash now resolves to effect_identity_model.v2_7.output.txt (as v2.6's did to .v2_6.output.txt in v2.7). A historical package therefore no longer self-verifies that one witness by path; the bytes are retained under the versioned name and the v2.8 crosswalk records both hashes in history_note.
A validator's CLI block ran on import. The receipt tool imported check() and the validator parsed its argv, trying to open a file named make. A module that is both a library and a CLI must guard the CLI with an entry-point check.
A subclass of a frozen object cannot add own properties (World freezes its instances); module-private state keyed by a WeakMap is the honest "inside the boundary" representation.
Audit the anchors before the run: walking the build with ast and counting every literal / regex anchor in the v2.7 documents found the only mismatch in one pass — one refusal per run is the expensive way.
A mutation must run where its defect is reachable: the v2.7-rule mutation on STRICT + NAIVE showed nothing, because that reconciler never learns satisfaction; on KEY-ONLY + NAIVE it reintroduced the avoidable duplicate. The same fact became an expectation ("revalidation needs a far side that can say satisfied").
The gate caught five phrases my own edits left (two version-history sentences still calling F36/F37 unadjudicated; "candidate for PROVED"; the typo phrase quoted verbatim, which is itself the forbidden phrase; one promotions without the rename wording). Fixed in the build, never by hand in the package.
F38 / NC31 adjudication and the R0.8 freeze decision (yours); the TRVM registration and the Super implementation (their lanes; drafts delivered); minting anything in the factory's canonical ref; TLC; EMB-E1; Graphonomous MCP not called (hang history).
inputs-gpt-execution-adjudication-v4.md)package-v2.7/ supersedes v2.6 (structure ACCEPTED; evidence state patched here) — an evidence-state patch (v4 §5, §16.A). diff -r package-v2.6 package-v2.7 is the delta; derived by tools-lane/build_v2_7.py (every anchor exactly once; every receipt hash recomputed; every expectation count derived by running) plus tools-lane/drift_gate_v2_7.py. The v2.7 gate PASSES on v2.7, FAILS on v2.6 (54 findings) and FAILS under twenty-three in-memory mutations (tools-lane/gate-v2_7-negative-control.txt).
| record | v2.6 | v2.7 | adjudication |
|---|---|---|---|
| E-13b | CLOSED-F31 / CLOSED-F32; F35 not covered | + CLOSED-F35 under the single-host deployment-anchor profile; the middle boxes (F36/F37, R0.8.7) stated as repaired-unadjudicated; typed witness = the R0.8.6 receipt (NC28 FAIL→PASS), pre-fix + repair receipts, subject_identity 747f025, adjudication_ref v4 §1–§2 | v4 §1–§2 |
r0_8 | F35 OPEN-UNADJUDICATED; power-loss among open findings | `profile` explicit (process crash / restart; single host; trusted root+anchor directory; content-addressed successor head; NO power-loss / consensus theorem); f35_status CLOSED; `nc29_status` EXECUTED (R0.8.7: F36 + F37 reproduced then repaired, 21/21, shipped whole, UNADJUDICATED); nc29_requirements_map a–e (e not visible); open_findings = F36/F37 repaired-unadjudicated only; `out_of_scope_profile` = power-loss, post-rename durability, consensus; freeze_decision PENDING | v4 §3–§4, §16.B–C |
| E-50a | one rule (unresolved or satisfied) | two rules (UNKNOWN → REPLAY_RISK_AUTHORIZED; SATISFIED → refused as done, a repeat is a NEW intent; FAILED-absent → free); adjudication_ref v4 §9 | v4 §9 |
| E-51 | 14/14 | re-executed v2.7 (19/19): SATISFIED-REPEAT unreachable under J1 at claim time; a repeat under a new intent K2 lands once; the v2.5 and v2.6 rules kept as mutations; the v2.6 receipt kept append-only (effect_identity_model.v2_6.output.txt) | v4 §9–§11 |
| L1? | ENV-SUCCESSOR + SYS-ATTACH-FAIRNESS (19/19) | three assumptions — ENV-SUCCESSOR-AVAILABILITY, SYS-SUCCESSOR-ENABLEMENT, SYS-ATTACH-FAIRNESS (21/21; v2.6's "both" now fails); statement_temporal_v2_7; TLA+ twin updated, unexecuted | v4 §12 |
promotions[] | typed, hashed, executed | + `from_token`/`to_token` beside prose, `subject_identity`, `adjudication_ref`/`adjudicator`/`adjudicated_at` where cross-lane, cross_lane; receipt hashes recomputed at build | v4 §6 A–C |
projection (new) | — | evidence_state.json GENERATED by the build; its sha256 recorded here and re-checked by the gate | v4 §6 D |
factory_candidates | FAC-CONTROL-SENSITIVITY "ready" | DECLARED (§8) + station_rule_v4 (§7) + schema_additions_v4; `FAC-EVIDENCE-TRANSITION-SCHEMA-CLOSED` drafted (candidate PROVED / exhaustive-in-declared-domain) | v4 §6–§8 |
frontier: header; S3 evidence (F35 closed; R0.8.7 executed NC29 → F36/F37); S5 row (E-50a two rules); §5 item 8; §7 (three assumptions; the v2.7 witness); §8 (DECLARED, the narrow claim, the station rule); synthesis / experiments / handoff / README correspondingly.
handoffs/SUPER_E50_EFFECT_IDENTITY.md v5 addendum (§9–§11); handoffs/TRVM_E42_E43_CANDIDATE_LAWS.md v5 addendum (§13–§14); tools-lane/factory-kit/ (rules A'/B'/C', fourteen mutations, the second claim draft, the station rule).
px13/ ships `COMPUTEDRIVEN_R087_FOR_GPT.{md,tar.gz}` whole (5.4 MB; commit 21a1452; internal manifest verified; series 21/21); px13/README.md updated; SHIPPED-SHA256SUMS.txt regenerated last.
package-v2.7-staging/ (the v2.7 witnesses as authored).
F36/F37 adjudication and the freeze decision (yours); TRVM registration; the Super-lane repairs; minting anything in the factory; TLC; EMB-E1.
inputs-gpt-execution-adjudication-v3.md)package-v2.6/ supersedes v2.5 (ACCEPTED with an adjudication patch) as the candidate canonical package — an evidence-state patch, not a research round (v3 §18). diff -r package-v2.5 package-v2.6 is the delta; derived by tools-lane/build_v2_6.py (every anchor exactly once; every receipt hash computed, every expectation count derived by running the witness) plus tools-lane/drift_gate_v2_6.py. The v2.6 gate PASSES on v2.6, FAILS on v2.5 and FAILS under thirteen in-memory mutations (tools-lane/gate-v2_6-negative-control.txt).
| record | v2.5 | v2.6 | adjudication |
|---|---|---|---|
| E-13b | TESTED-CONDITIONAL / CLOSED-F31 | + CLOSED-F32 (the anchor pins revision + head) under the single-host deployment-anchor profile; F35 explicitly not covered (one ahead proves ordering, not provenance); typed witness NC26 FAIL→PASS; pre-fix (R0.8.4) + repair (R0.8.5) receipts hashed | v3 §1–§2 |
| E-14 | TESTED-CONDITIONAL (F23/F24) | + F33 (the intended interpreter verified, not the observed one committed), F34 (integrity_check), F24 properly (ordering in Rust, duplicates refused); committed / not_claimed updated | v3 §3–§5 |
r0_8 | open findings incl. "F22 layer", "F30 pending" | closed_by_adjudication_v3 (F31, F23/F24, F32, F33, F34, F24, F22, F30); `f35_status = OPEN-UNADJUDICATED`; open findings = F35 + power-loss + post-rename durability + single host + the filesystem namespace as a trust assumption; F22/F30 "pending" removed (the gate forbids them) | v3 §6–§8 |
| L1? | witnesses with fairness fields; "unresolvable" used for eventuality failures | vocabulary corrected (eventuality-failing; resolvability checked beside eventuality; UNRESOLVABLE only for NO-FINALITY); ENV-SUCCESSOR / SYS-ATTACH-FAIRNESS split executed (offer/attach refinement; 19/19); statement_temporal_v2_6; witness sha256 + executed + runtime on every entry; TLA+ twin updated, still not run | v3 §9–§11 |
| E-48 | TESTED (profile in prose) | TESTED under the exact stated profile — tested / not_tested lists; executed + runtime + receipts (token unchanged: not a promotion) | v3 §12 |
| E-50 | one record | split: E-50a replay-risk authorization by stable intent identity (REPLAY_RISK_AUTHORIZED, pre-outcome, not a terminal) · E-50b attempt supersession / reconciliation attribution (SUPERSEDED; attempt id in the far-side record); both FALSIFIED-KEPT-RED for the Super lane | v3 §13 |
| E-51 (new) | — | S5 / representation — EXHAUSTIVE-IN-MODEL (14/14): the effect-identity model over EXP-6's shape; the key-state rule; the strict reconciler; the model's own first rule kept as a mutation | v3 §13, §21.4 |
| E-49 | TESTED as adapters (22/22) | + A-IDEMPOTENT and L-ANCESTRY (35/35) — token unchanged | v3 §15–§16 |
promotions[] | prose sensitivity witnesses | typed (type, receipt, sha256, executed, runtime; pre_fix_witness + repair_witness where the history includes FALSIFIED); every record carries evidence_class_token (rule A compares tokens, so text refinements are not promotions) | v3 §17 |
evidence_state.json (new) | — | the generic evidence-state@1 projection the factory kit validates | v3 §17 |
factory_candidates | candidate | ready for a factory round; generic_rules_v3 A–H; tools-lane/factory-kit/ (validator + selftest + draft claim + round steps) | v3 §17 |
frontier: header; S3 evidence (R0.8.5 adjudicated, F35 open, R0.8.6 shipped); S5 row (E-50a/b, E-51, the exact profile); §5 items 6–8; §6 count 56; §7 (L1-service-succession block; the v2.6 witness paragraph); §8 (rules A–H; the kit);
synthesis: title; S3 / S5 rows; §4 E; §6 rows 1, 6, 7; §7; §8; experiments: title, rows 1, 6, 7, L1?; handoff: title, item 1, B/C/E/F/G; README: title, paragraph, rows (crosswalk, evidence_state.json, fair witness, inputs), reproduce line.
px13/ ships `COMPUTEDRIVEN_R086_FOR_GPT.{md,tar.gz}` whole (4.8 MB; internal manifest verified; series 21/21 at 747f025); px13/README.md v4 paragraph; SHIPPED-SHA256SUMS.txt regenerated last.
handoffs/SUPER_E50_EFFECT_IDENTITY.md (new); tools-lane/factory-kit/ (new); package-v2.6-staging/ (the v2.6 witnesses as authored).
F35 adjudication (yours, from the shipped handback); TRVM registration (the lane chooses A-idempotent / B and the fork semantics); the Super-lane repairs E-50a/E-50b name; minting FAC-CONTROL-SENSITIVITY (the factory round); TLC; EMB-E1.
inputs-gpt-execution-adjudication-v2.md)package-v2.5/ supersedes v2.4 (ACCEPTED as canonical by the v2 adjudication) as the candidate canonical package. diff -r package-v2.4 package-v2.5 is the delta; it was derived by tools-lane/build_v2_5.py (every anchor exactly once) plus tools-lane/drift_gate_v2_5.py. The v2.5 gate PASSES on v2.5 and FAILS on v2.4 and under six in-memory mutations of the v2.5 evidence state (tools-lane/gate-v2_5-negative-control.txt) — the gate now checks evidence-state facts, not only prose.
| record | v2.4 | v2.5 | adjudication |
|---|---|---|---|
| E-13b | FALSIFIED at efa8881 (F31), repaired in R0.8.4, closure to adjudicate | `TESTED-CONDITIONAL / CLOSED-F31` — closed under {single host; one protected deployment-owned anchor namespace; one LocusId → one incarnation lock}; the anchor namespace is part of the trust profile (F22 layer); conditions, trust_profile, sensitivity_witness (NC24 FAIL→PASS) | v2 §2 |
| E-14 | TESTED-CONDITIONAL / FALSIFIED-OPEN | `TESTED-CONDITIONAL` — F23/F24 closed under a trusted/fixed SQLite interpreter with the schema interpretation contract committed; committed, not_claimed, conditions; FALSIFIED-OPEN removed for those attacks | v2 §3 |
| E-15 | live_local under a fixed / trusted schema and interpreter | `TESTED` under committed schema + trusted interpreter + defensive configuration (profile qualifier retained) | v2 §4 |
r0_8 | a status string | `OPEN` + `open_findings[]` (power-loss UNCLAIMED; directory-fsync post-rename semantics; single host; the F22 trust boundary as an assumption; F30 adjudication pending) + the lane's R0.8.5 / R0.8.6 reports, closure to adjudicate | v2 §5 |
| L1? | one witness, described as liveness | two witnesses with explicit `fairness` / `temporal_eventuality`: liveness_l1.py relabelled L1-PRE / resolvability (fairness: none); `liveness_l1_fair.py` the temporal witness under a declared policy (WF(finality); WF(attach) for the successor clause; 15/15; two algorithms agree; three self-mutations flip); L1Succession.tla + .cfg for TLC (not run, no JVM); statement_temporal_v2_5; finding: clause A is an environment-liveness assumption | v2 §7 |
| — | — | `factory_candidates.FAC-CONTROL-SENSITIVITY` (factory epistemic / evidence-admission rule; witness families; handoffs/FAC_CONTROL_SENSITIVITY_CANDIDATE.md); frontier §8, synthesis §8 | v2 §6 |
| E-48 (new) | — | S5 / mechanism — TESTED: EXP-6, the first real adapter run (ampd Gateway.perform/5 + a keyed git far side; naive duplicates, keyed deduplicates; UNKNOWN live under process kill and BEAM SIGKILL; reconciliation attempt-aware) | v2 §12.6 brought forward |
| E-50 (new) | — | S5 / corollary — FALSIFIED-KEPT-RED for the super lane: journal-level effect identity (a same-key re-drive under fresh authority admitted under a new id; no SUPERSEDED / DUPLICATED_ACKNOWLEDGED terminal) | EXP-6 |
| E-49 (new) | — | S4 / mechanism — TESTED as adapters: EXP-7, E-43 contracts A/B and E-42 L-PREFIX/L-ID over the pinned trvm_world.mjs, 22/22, costs measured; E-42/E-43 stay red | v2 §8 |
| ledger P5 | SUPPORTED | TESTED (EXP-6, scoped profile); C-4 gains a live witness | EXP-6 |
promotions[] | — | E-13b, E-14, E-15, E-48 — each with a sensitivity_witness naming a receipt | v2 §6/§9 |
frontier: header; S3 evidence cell (closures under profiles, open findings pointer); S5 row (EXP-6, profile); §5 items 6–7; §6 count 54; §7 temporal form of L1 + the two witnesses; §8 factory epistemics (new);
synthesis: title; S2/S3/S5 rows; §4 B and E; §6 rows 1, 6, 7; §7; §8 (new);
experiments: title; status rows 1, 6, 7, L1?; handoff: title, do-not-change (production adapters), item 1, the S5 test bullet, the next-agent sequence A–G; README: title, paragraph, rows (crosswalk, three witnesses, inputs), reproduce line (the v2.4 negative control), receipts line;
EXTERNAL_RESEARCH_PINS_V3.md (new): the eight references the v2.5 additions lean on, fetched this round (Alpern–Schneider; Lamport ch. 8; ProB fairness; Kleppmann; Stripe; Two Generals; DeMillo–Lipton–Sayward; Barr et al.);
historical round records keep their original sentences inside <!-- PROVENANCE --> blocks.
px13/ ships `COMPUTEDRIVEN_R085_FOR_GPT.{md,tar.gz}` whole; px13/README.md carries the v3 paragraph; `px13/SHIPPED-SHA256SUMS.txt` regenerated AFTER the README edit (the v2 blemish); R0.8.6 quoted, not shipped (no bundle existed).
new experiment directories: experiments/s5_effect_adapter/ (ampd vendored from the committed object super 7651697; two receipt sets — run 1 key-only reconcile kept, run 2 attempt-aware) and experiments/trvm_contracts/.
R0.8.5 / R0.8.6 closure adjudication (computedriven lane's receipts; nothing re-run here); TRVM registration (the lane chooses a contract); the super-lane repairs E-50 names; EMB-E1 (operator spend); TLC on the TLA+ twin; minting FAC-CONTROL-SENSITIVITY in the factory.
inputs-gpt-execution-adjudication.md)package-v2.4/ supersedes v2.3 as the candidate canonical package. diff -r package-v2.3 package-v2.4 is the delta; it was derived by tools-lane/build_v2_4.py (every anchor exactly once) plus tools-lane/drift_gate_v2_4.py. The v2.4 gate PASSES on v2.4 and FAILS on v2.3 (negative control in tools-lane/gate-negative-control.txt).
| record | v2.3 | v2.4 | adjudication |
|---|---|---|---|
| S6? | resolved in prose; S6? still a vocabulary value | retired as an obligation value; resolved_candidates.S6?; README/synthesis/handoff say five obligations + S1 corollaries | §1 |
| — | — | `liveness_candidates.L1?` recoverable succession / explicit terminalization, with witnesses/liveness_l1.py (6/6); frontier §7, synthesis §7 | §2 |
| E-13 | live_local (narrowed) | split: E-13a same-anchor exclusivity live_local (TESTED) · E-13b canonical host-local anchor `FALSIFIED at efa8881` (F31), repaired in R0.8.4, closure to adjudicate from the shipped handback · E-13c witnessed multi-namespace currentness TESTED under the ARM-B profile | §4 |
| E-14 | live_local (narrowed) | `TESTED-CONDITIONAL` (fixed/trusted SQLite interpretation) / `FALSIFIED-OPEN` at efa8881 (F23/F24); R0.8.4 repair to adjudicate | §4 |
| E-15 | live_local | live_local under a fixed/trusted schema and interpreter (F23 trigger case) | §4 |
| E-42 | S3 / corollary | S4 / representation, S3 by cross-reference; → TRVM lane | §6 |
| E-43 | S4 / corollary | unchanged; → TRVM lane (handoffs/TRVM_E42_E43_CANDIDATE_LAWS.md) | §6 |
| E-46 | one S2 record | split: E-46a CROSS/representation (join-safety; the conjunction case is a merge-contract question, not amplification evidence) · E-46b S2 (union resurrects revoked/spent authority) · E-46c S1 (stale-true action at the target) | §7 |
| S3 scope text | "cross-host … TESTED" | exact profile: TESTED on a two-controller / two-filesystem-namespace copied world under a single correct, available witness; SUPPORTED by mechanism shape (cross-machine, same assumptions); UNTESTED (partition-/failure-tolerant witnessing) | §3 |
| S4 scope text | TESTED (holds) | unchanged | — |
r0_8_status | — | OPEN: R0.8.3 executed (historical); R0.8.4 falsify-then-repair of F31/F23 + F25–F29; F22/F30 deferred; R0.8.5 in flight, not consumed | §5 |
every primary table/status moved from pre-execution to executed (synthesis §1 rows, §4 A/B/E, §6; frontier §1 rows, §4 stale paragraph removed, §5 statuses; handoff opening + items 1–5 + next-agent sequence; experiments items 1–5 + status table); crosswalk count 41 → 51 stated where counted;
drift gate v2.4 asserts post-execution statuses (S6 TESTED + S1 corollary + not an obligation; README count == JSON count; R0.8 OPEN + EXECUTED; single-witness/two-controller wording present and cross-host/fault-tolerant TESTED absent; L1 on its own axis; splits present);
historical round records (FABLE51_R10PRE_*) keep their original sentences inside <!-- PROVENANCE --> blocks.
top-level SHA256SUMS no longer lists itself; the archive hash is published beside the zip (.sha256 sidecar) and in the return message;
px13/ now ships the complete original handbacks (COMPUTEDRIVEN_R083_EXECUTION_FOR_GPT.tar.gz, COMPUTEDRIVEN_R084_FOR_GPT.{md,tar.gz}) and SHIPPED-SHA256SUMS.txt covering exactly what ships; the old handback manifest is kept as R083-handback-internal-SHA256SUMS.txt (it describes the tarball's contents);
EXP-3 and the ampd half of EXP-5 now run from vendored pinned artifacts (experiments/observe_is_write/vendor/, experiments/authority_merge/vendor/, each with SHA256SUMS + PROVENANCE.txt); env overrides still point them at a live tree.
R0.8 repairs / re-execution (§12 B): the computedriven lane's; its R0.8.4 result is shipped, not re-run; R0.8.5 in flight.
TRVM E-42/E-43 (§12 C): handed off, not registered.
EMB-E1 (§12 D), S5 adapter (§12 E): unchanged.
Package package-v2.2/ is the handoff as received (GPT-5.6 EXECUTION GO, 2026-09-01). package-v2.3/ is the same package with the execution results folded in; diff -r package-v2.2 package-v2.3 is the whole delta. The drift gate prints DRIFT GATE PASS on v2.3 (package-v2.3/tools/drift_gate.output.txt); the v2.2 witness is untouched (sha256 439ed40b…, 19/19, stdout ad23cf4b… across four hash seeds).
| claim | from | to | receipt |
|---|---|---|---|
| S6? world/locus birth uniqueness (ledger C-7; frontier §1 row, §4) | UNKNOWN / candidate / "predicted reachable" | TESTED reachable, both act, cross-birth finality admitted → recorded as a corollary: S1(domain = world locus-birth) | experiments/rust/results/S6_locus_birth.md, crosswalk E-44 (new) |
| E-12 recovery is re-derivation (S3) | in_tree | live_local | px13/ (R0.8.3 21/21 @ efa8881, nc7 rows, NC15) |
| E-13 exclusivity outside the copyable store (S3 mechanism) | in_tree, "single host (OS lock)" | live_local (narrowed): one lock namespace only — F31; cross-host: FORK without anchor (TESTED), ONE advance with a witnessed root (TESTED, witness trust profile) | px13/ nc1/nc6; experiments/distributed_currentness/receipts/EXP1-*.json |
| E-14 root commitment under injective encoding (S3 mechanism) | in_tree | live_local (narrowed): SQLite interpretation contract trusted (F23/F24 open) | px13/ nc2/nc4/nc8 |
| E-15 mutation-time canonicality (S3 mechanism) | in_tree | live_local | px13/nc17-18-19-23-live-tamper.txt (24 gates + verdict) |
| S4 scope note "observation-as-write UNTESTED (EXP-3)" (frontier §1; E-22 annotated) | UNTESTED | TESTED — holds on live v0.12.0 | experiments/observe_is_write/ (15/15) |
| S3 scope note "cross-host UNKNOWN (EXP-1 ARM B)" (frontier §1) | UNKNOWN | fork without anchor / one advance refused by name and quarantined with one witnessed root | experiments/distributed_currentness/ |
| C-5 authority merge (reworded claim) | EXHAUSTIVE-IN-MODEL | + TESTED on two live APIs (cd-core reconstruct, Ampd.Core.snapshot_of) | crosswalk E-46 (new); experiments/authority_merge/ |
| C-2 Locus (narrowed statement) | NARROWED | unchanged — census added: 5 / 4 / 18 / 0 of 27; residue enumerated | crosswalk E-45 (new); experiments/locus_capability_facade/ |
| C-1 dual anchors as the independence witness | EXHAUSTIVE-IN-MODEL | unchanged — per-substrate witness added, mechanism count shown invisible to the model | crosswalk E-47 (new); experiments/micro_locus_s3/ |
| record | what stays red | for whom |
|---|---|---|
| E-42 | a world fork issues the same versions as its parent for different values; verification across the fork boundary splits the verifiers | TRVM lane — candidate warrant.lineage-bound@1; S3 at the world layer |
| E-43 | registerQuery is write-guarded but not version-advancing; two worlds identical under (vclock, log, resources) observe differently | TRVM lane — candidate world.scope-registry-versioned@1 |
S1, S2, S4, S5 statements: no falsification. S5 still has no adapter run.
EMB-E1: NOT executed (operator spend protocol); emb_e1/STATUS.md.
Nothing promoted to an irreducible basis. The finality/liveness dimension surfaced by EXP-4 is recorded as a gap in experiments/locus_capability_facade/RESULT.md, not as a new S-row.
EXP-5's first-run expectation was mis-specified against its own measurement (the set union collapses the re-grant in every order: the verdict flips, it does not vary). Corrected; first run kept beside the result.
The live computedriven/ tree moved (efa8881 → R0.8.4) while this lane ran; nothing was consumed from it — every build is from git archive efa8881 (experiments/vendor/, SHA256SUMS).
The v2.2 package said the PX13 lab was not executed; it had been (2026-09-01, by the computedriven session). Folded in as evidence with its caveats (F22–F25, F31), not re-run.