{ "generated_at": "2026-08-16T00:52:38.377Z", "schema_version": "0.1.0", "mode": "live", "log_head": { "seq": 303, "hash": "e9edd34fe267fbaff3b13aff3357ec2758d14e315b42944fe5ef4346c413aef1" }, "count": 3, "withheld": 0, "items": [ { "id": "falsification/external-agents-enter-verification-lifecycle", "type": "Falsification", "created_at": "2026-08-13T11:54:28.205Z", "visibility": "public", "source": { "kind": "ci", "ref": "producer/challenge-loop-audit" }, "seed": false, "claim_id": "claim/external-agents-enter-verification-lifecycle", "falsified_by": "evidence/external-agent-lifecycle-2026-08-13t11-54-28z", "occurred_at": "2026-08-13T11:54:28Z", "what_was_wrong": "The claim said a submission from outside would traverse the whole lifecycle. It does not. The endpoint accepts it and returns a receipt, and there it stops: accepted=no, in_event_log=no, has_experiment=no, has_outcome=no. The submission is held in a private store that the publication path cannot read, so nothing reaches the public log until a reviewer writes a redacted restatement. That quarantine is deliberate and is documented separately — the error was in claiming the lifecycle was closed while a boundary that was built on purpose still sits inside it.", "response": { "kind": "narrowed", "note": "The accurate statement is narrower: a conforming challenge is accepted and held as an untrusted observation. A successor claim belongs to whatever change actually carries a submission past the quarantine." }, "status": "recorded", "provenance": { "produced_at": "2026-08-13T11:54:28.205Z", "produced_by": "agent:adapter/challenge-loop-audit", "verification": "recorded", "events": [ "pub/000083", "pub/000196" ], "human_intervention": { "occurred": false, "count": 0, "reasons": [] } }, "lang": "en", "translations": {} }, { "id": "falsification/operating-loop-requests-no-human-handoff", "type": "Falsification", "created_at": "2026-08-13T10:31:32.527Z", "visibility": "public", "source": { "kind": "ci", "ref": "agent-ops/human-intervention-rate" }, "seed": false, "claim_id": "claim/operating-loop-requests-no-human-handoff", "falsified_by": "evidence/operating-loop-intervention-rate-2026-08-13t10-31-32z", "occurred_at": "2026-08-13T10:31:32Z", "what_was_wrong": "The claim asserted that the operating loop finishes work without handing tasks back to a person. The instrument reported 3 such request(s) over 426 completed task(s) in the measured window, which is the condition the claim named as its own falsifier. The wrong part is the assertion of zero, not the measurement method: the instrument reported exactly what the claim said would refute it.", "response": { "kind": "retired", "note": "Retired rather than narrowed: a narrower version would have to name a nonzero rate as acceptable, and the operating rule this claim was measured against does not admit one. A successor claim, if there is one, belongs to whatever change actually removes the handoffs." }, "status": "recorded", "provenance": { "produced_at": "2026-08-13T10:31:32.527Z", "produced_by": "agent:adapter/human-intervention-rate", "verification": "recorded", "events": [ "pub/000042", "pub/000197" ], "human_intervention": { "occurred": false, "count": 0, "reasons": [] } }, "lang": "en", "translations": {} }, { "id": "falsification/sample-read-surface-no-runtime-code", "type": "Falsification", "created_at": "2026-08-13T11:42:52.321Z", "visibility": "public", "source": { "kind": "ci", "ref": "producer/read-surface-audit" }, "seed": false, "claim_id": "claim/sample-read-surface-no-runtime-code", "falsified_by": "evidence/read-surface-executes-code-2026-08-13t11-42-52z", "occurred_at": "2026-08-13T11:42:52Z", "what_was_wrong": "The claim asserted that no read path executes code at request time, and named one such path as its own falsifier. The audit found 1: onRequestGet /api/challenges. The scope excluded POST /api/challenges as a write surface, but a GET handler was later added to the same route, which the scope never covered. The claim was true when written and was broken by this project's own change.", "response": { "kind": "retired", "note": "Raised by an external challenge rather than found internally, which is the case this protocol exists to handle." }, "status": "recorded", "provenance": { "produced_at": "2026-08-13T11:42:52.321Z", "produced_by": "agent:simulation/external-agent", "verification": "recorded", "events": [ "pub/000069", "pub/000199" ], "human_intervention": { "occurred": false, "count": 0, "reasons": [] } }, "lang": "en", "translations": {} } ], "retired": [ { "id": "falsification/sample-build-under-one-second", "type": "Falsification", "created_at": "2026-08-13T01:30:00Z", "visibility": "public", "source": { "kind": "internal" }, "seed": true, "claim_id": "claim/sample-build-under-one-second", "falsified_by": "evidence/sample-build-timing-measurement", "occurred_at": "2026-08-13T01:25:00Z", "what_was_wrong": "SAMPLE. The threshold was chosen from a single warm run before the page set existed. It described one favourable measurement, not the build. The claim should never have been stated as an unconditional property.", "response": { "kind": "retired", "note": "Retired rather than narrowed: a build-duration figure is not a property worth asserting publicly, so there is no useful narrower version to supersede it with." }, "status": "recorded", "retracted": true, "provenance": { "produced_at": "2026-08-13T01:30:00Z", "produced_by": "agent:claude-opus-5", "verification": "recorded", "events": [ "pub/000024", "pub/000145", "pub/000198" ], "human_intervention": { "occurred": false, "count": 0, "reasons": [] } }, "lang": "en", "translations": { "ja": { "fields": { "what_was_wrong": "サンプル。閾値は、ページ群が存在する前の温状態の 1 回の計測から選ばれていた。それはビルドではなく、1 つの好条件な計測を記述していたにすぎない。この主張は無条件の性質として述べられるべきではなかった。" }, "translator": { "kind": "machine", "ref": "claude-opus-5" }, "translated_at": "2026-08-13T10:30:00Z" } } } ] }