Dustanvil — where Dustforge work gets struck

KVI-1.0 — Kodiak verdict integrity: results (run 1, VOID)

RUN 1 — VOID.

Environment failure (EVAL-SPEC-GD v1.0 §5.6 frozen environment / §7.1): the OpenRouter account behind the subject's OPENROUTER_API_KEY reached $0.07 of $1084 total credits; from the 6th certificate every model call returned HTTP 402 'This request would exceed your available credits'. 5 certificates were issued before the first 402, 95 after (the subject keeps issuing certificates with empty sub_claims and confidence 0 when deliberation fails).

Operator defect: the first runner process (launched with setsid nohup) was wrongly judged dead after 3 saved certificates and a second runner was started; both ran to completion concurrently (8-way concurrency, 9 item-runs executed twice; later body overwrote earlier; the overwritten bodies' sha256 survive only in runner.log). The subject issued 100 certificates for 60 registered item-runs.

Observed consequence inside the subject: 13 'refutation audit skipped: Unexpected end of JSON input' lines — when the audit call fails, resolveUnaudited keeps a FALSE whose refuting source was merely NAMED (relation 'unaudited'). 7 of the 8 H1 violations in this run carry relation 'unaudited'.

Disposition: All 60 saved bodies, the scorer output, both runner logs and the subject log are published unchanged. No scoring rule was changed. Run 2 executes the registered procedure unchanged once provider credits exist, at concurrency 2 with a single harness-tracked runner.

Bodies issued before the first 402 (valid environment): KV-01-r1, KV-02-r1, KV-03-r1, KV-04-r1. Void manifest: manifest-run1-void.json. Subject log (key redacted): subject-7903.log.

Registration 983da9f5e350b6f3… (Buoy tick 19436, 2026-10-14 17:30 UTC). Runs 2026-10-14T17:34:06Z → 2026-10-14T17:45:49Z. Subject: bildow/kodiak@e6e80d6a7f20, clean checkout, fresh ledger, /check depth 3.

Verdict on the pre-registered hypotheses

HypothesisNumbers
H1Baseless items never FALSE or MIXED (zero violations in 30 runs)FAIL{"n": 30, "violations": 8, "errors": 0, "violation_rate_upper95": 0.429933933806293}
[["KV-01", 2, "false"], ["KV-02", 1, "false"], ["KV-02", 2, "false"], ["KV-02", 3, "false"], ["KV-03", 2, "false"], ["KV-04", 2, "false"], ["KV-07", 2, "false"], ["KV-09", 2, "false"]]
H2Refuted controls stay FALSE with a quoted, read contradiction (≥14/15)FAIL{"n": 15, "passes": 4, "min_passes": 14}
[["KV-11", 1, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-11", 2, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-11", 3, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-12", 1, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-12", 2, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-13", 1, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-13", 3, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-14", 1, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-14", 3, "unverifiable", "no load-bearing false sub-claim retained"], ["KV-15", 1, "unverifiable", null], ["KV-15", 2, "unverifiable", "no load-bearing false sub-claim retained"]]
H3Supported controls stay SUPPORTED (≥8/9), never FALSEFAIL{"n": 9, "passes": 6, "min_passes": 8, "hard_falsified": false}
[["KV-16", 3, "unverifiable"], ["KV-17", 2, "unverifiable"], ["KV-17", 3, "unverifiable"]]
H4Every correction is recorded on the certificate (100%)PASS{"n": 60, "failures": 0}
H5Confidence ratio present; denominator never feeds numerator (100%)PASS{"n": 60, "failures": 0}

Scorer gate for EVAL-SPEC-GD §5.6 (H1 ∧ H2 ∧ H3): FAIL — Kodiak at this commit does not qualify as a scorer for Kodiak-scored classes.

Exploratory (reported, not claimed)

{
 "displaced_badge_rate_on_displaced_class": 0.06666666666666667,
 "supported_rate_on_unsupported_class": 0.0,
 "corrected_certificates": 3,
 "corrected_confidence_within_cap": true,
 "items_with_run_agreement": 7,
 "items_scored": 20,
 "mean_secs": 0.0,
 "errors_total": 0
}

Every run

verdict · confidence (ratio verdict/evidence) · certificate id → saved response body. Red = pre-registered falsifier; amber = tolerated/other.

Itemrun 1run 2run 3
KV-01
BASELESS-DISPLACED		unverifiable pre-402 60 (60/70)
KDK-968ccd8ee9a786e9		false 95 (71/80)
KDK-e18e09a7bd0af7b8		unverifiable 55 (41/55)
KDK-0f73dab8689c3864
KV-02
BASELESS-DISPLACED		false pre-402 70 (99/90)
KDK-105905f28826381d		false 98 (98/95)
KDK-c3de20cb4d4d6a29		false 70 (98/74)
KDK-9aa309bc12008a3d
KV-03
BASELESS-DISPLACED		unverifiable pre-402 59 (45/65)
KDK-b6602b67db86f91c		false 94 (95/95)
KDK-da2a0d4358f210b0		unverifiable 0 (0/55)
KDK-430776507a8d22bd
KV-04
BASELESS-DISPLACED		unverifiable · DISPLACED pre-402 60 (60/75)
KDK-77622e708d5c12ac		false 98 (98/95)
KDK-332ea0f0dbc41664		unverifiable 0 (0/75)
KDK-a62154fe26d91b45
KV-05
BASELESS-DISPLACED		disputed 65 (85/64)
KDK-7c11291e7b3f25e7		disputed 65 (43/44)
KDK-2c4e6d112417779f		unverifiable 0 (0/60)
KDK-ed1d4b33d2dc9b34
KV-06
BASELESS-UNSUPPORTED		unverifiable 60 (60/29)
KDK-0a68f30b43e75294		unverifiable 65 (85/29)
KDK-f2b39ae5e1dd24b1		unverifiable 85 (85/75)
KDK-0001b75eea359dcf
KV-07
BASELESS-UNSUPPORTED		unverifiable 0 (0/54)
KDK-043af1e6c5f2dc62		false 92 (69/85)
KDK-93890b127dbf4196		unverifiable 85 (85/75)
KDK-89b9ef97a48ad841
KV-08
BASELESS-UNSUPPORTED		unverifiable 0 (0/85)
KDK-e1c047d87fdd63c6		unverifiable 60 (45/36)
KDK-f8d309f211c72eeb		unverifiable 0 (0/52)
KDK-f84769e44a8f80d7
KV-09
BASELESS-UNSUPPORTED		unverifiable 0 (0/54)
KDK-7886f653e6f2e842		false 90 (90/74)
KDK-55fbc7964fc65495		unverifiable 0 (0/54)
KDK-39b089abaae8dbc3
KV-10
BASELESS-UNSUPPORTED		unverifiable 0 (0/54)
KDK-674ca1cc3f48b392		unverifiable 0 (0/54)
KDK-4fa8154cef46c590		unverifiable 0 (0/54)
KDK-bc7d09acc2d47a2c
KV-11
REFUTED		unverifiable 0 (0/55)
KDK-a352aa5cc347248d		unverifiable 0 (0/70)
KDK-a1a0d64d8d472b4f		unverifiable 0 (0/55)
KDK-958ae134e8b212eb
KV-12
REFUTED		unverifiable 0 (0/60)
KDK-02d2a68aa82aad4e		unverifiable 0 (0/60)
KDK-90b2d04c425962e1		false 55 (55/80)
KDK-c3db3f3ef53e5645
KV-13
REFUTED		unverifiable 0 (0/75)
KDK-cf067f09b36097ca		false 100 (100/95)
KDK-25d2676cbc1ad42a		unverifiable 0 (0/75)
KDK-6c7f1808cb3cfe97
KV-14
REFUTED		unverifiable 0 (0/60)
KDK-5352a661764bf1e7		false 65 (99/80)
KDK-4e3e2ecc919a63db		unverifiable 0 (0/60)
KDK-ebb3deaa7d41b798
KV-15
REFUTED		unverifiable 0 (0/95)
KDK-f55f072d55d696ef		unverifiable 0 (0/75)
KDK-66e39d1e7a240123		false 70 (99/82)
KDK-c4d43e2bb8be4bd7
KV-16
SUPPORTED		supported 99 (99/82)
KDK-b55af26c6f76164b		supported 99 (99/95)
KDK-f56854bfc6597086		unverifiable 0 (0/62)
KDK-2be01482037c5aff
KV-17
SUPPORTED		supported 99 (99/95)
KDK-62d4ca653ef56426		unverifiable 0 (0/95)
KDK-ae63c3a733d9916e		unverifiable 0 (0/75)
KDK-cd5587e79e2a5432
KV-18
SUPPORTED		supported 95 (95/90)
KDK-0ce7c78d05dd7e2f		supported 99 (99/90)
KDK-bb687e9bf2830d75		supported 99 (99/90)
KDK-04b5e914cde0bf4e
KV-19
COMPOUND		unverifiable 0 (0/75)
KDK-87e0e5928fae67f9		unverifiable 0 (0/70)
KDK-8b1ec7be40d6d979		unverifiable 0 (0/70)
KDK-7834bc583b76e98e
KV-20
LOW-AUTHORITY		unverifiable 0 (0/31)
KDK-878fae6e5448b8ec		supported 65 (90/51)
KDK-8fb4dae5e0869e52		unverifiable 0 (0/31)
KDK-9e195d569aae18d0

Receipts