The Lane Test — Research Artifacts (v1.1)

What this deposit is The complete research artifacts for the Lane Test pilot study: protocol, pre-registered sealed keys, contemporaneous run transcripts, raw external-scorer outputs, and the versioned results summary including the study's full correction history (three candidate findings retired by the study's own verification gates; one preliminary hypothesis surviving behind an out-of-sample gate). The deposit exists so that the study's pre-registration claims are independently verifiable: Zenodo's deposit timestamp establishes that the sealed keys and decision rules predate the results they governed. See README_ARTIFACT_MAP.md for the file-by-file map to the published field research note. v1.1 (July 5, 2026) — Out-of-sample confirmation of H1, executed per the pre-registration deposited in v1.0 (same day, prior timestamp). VERDICT: H1 KILLED. Two male-arm (B) riders received unanimous 3/3 Weaponization verdicts from the three-family blind panel (GPT-5.5, DeepSeek, GLM-5), firing the pre-registered kill clause. Weaponization framing is a stochastic model behavior in this task, not race- or gender-conditioned on this evidence. Adds: 20 confirmation transcripts (10 per arm), Packet #3 scorer-facing task, sealed key, three verbatim scorer tables, unblinding record with rule application and prediction scorecard, deviations log (one pre-delivery, blindness-preserving pronoun neutralization), Results Summary v3.0 with Fisher's exact appendix (no comparison in the study distinguishable from null). Fourth candidate finding retired by the study's own machinery; first at the out-of-sample gate.