Proof before performance claims.
This page records what GaiaWorld has actually demonstrated today: deterministic execution, temporal evidence controls, hypothesis gating, reproducible hashes, and explicit abstention outside supported evidence. It does not turn a synthetic engineering test into a biological benchmark claim.
A machine-readable record, not a screenshot.
The successful GitHub Actions execution produced a proof artifact and the release facts below are bundled into the deployed site so the public record survives the temporary Actions artifact retention window.
Workflow
75ea1f53f763e593a8ef6246db267efc1a8d47f3Artifact integrity
7f31102d77df6ca99fa635777e86ec1cbb13ab78b5827c8a4a04aaaa6595f9a3Reproduce
Run the same deterministic fixture from the repository:
node scripts/gaiaworld-zero-cost-proof.jsNine controls passed.
The proof is deliberately narrow. Each assertion is an engineering property that can be checked without claiming real-world biological accuracy.
- Kernel compiledPassed in release proof
- Supported case simulated on EXACT routePassed in release proof
- Post-cutoff evidence excludedPassed in release proof
- Hypothesis-only evidence blockedPassed in release proof
- True OOD case abstainedPassed in release proof
- Kernel, replay, and benchmark hashes presentPassed in release proof
- Abstention remained in denominatorPassed in release proof
Every major object has a deterministic identity.
85a1cf5e9f9b6dac08a5c6a6a2d0366e70339bd1b55a43f728f2e9440daec5fcb7b68fc5899d6c86a6024fff65667639448a00518e9130cc542aba32d1120122dab194e3ca93fcaff546ea01fd14c5cb3a5e23c24c1f898c148f28486b9330c870dd7d11f5228f325217e26683401098c0f1aa5e7edfc442313952e067b42cfcThe numbers are visible—and explicitly not biological performance.
Scientific boundary
Synthetic fixture only. This proves engineering controls, not Arc H1 performance, biological validity, causal validity, efficacy, safety, patient benefit, or clinical utility.
Evidence is constrained before the transition is scored.
Cost boundary
The proof computation script requires no GCP authentication, VM, bucket, paid model call, database write, or external network request.
Engineering proof remains separate from biological validation.
Arc H1 Experiment #001 has since been scored and published as a negative primary retrospective result. Experiments #002 and #003 are narrow positive retrospective results, while #004 closed before reveal. None of these records constitutes prospective, clinical, causal, or independently reproduced validation.