What each check establishes.
A reproduced simulation, a calculation on saved measurements, and the provenance of an external request answer different questions.
These checks describe a synthetic learning world. They do not establish a law of nature or prove that IMD outperforms the local selector.
Reproduce a local world
Run the published simulator fixtures, make predictions and verify their revealed configurations.
What you can check
The offline verifier runs 24 separate local fixtures through four held-out conditions and checks each commitment/reveal. A completed free expedition also reveals its own configuration.
node verify.mjs
Where the claim stops
These fixtures are separate synthetic worlds. They do not rerun the historical IMD experiment, and 4/4 is not a measure of IMD intelligence.
Recalculate the recorded result
Use the original public measurements to check which candidate models are contradicted.
What you can check
Both recorded observations have growth 50/100, energy 40/100 and survival false. The water-only treatment rules out models 07, 11, 13, 14 and 16: 19 compatible models become 14.
node verify.mjs
Where the claim stops
The historical world’s hidden opening is not published. Its measurements are preserved inputs to this calculation; they cannot be independently regenerated from that hidden configuration.
Inspect the recorded measurements → · Download the canonical record → · Compare IMD and local choices →
Check external provenance
Match the paid IMD request to its admission, job and returned artifact using public Ethereum and IMD sources.
What you can check
The online verifier reads public RPC/API data and checks the request/event binding and artifact SHA-256. It also distinguishes the historical direct-Intake request from the separate GenesisResearch deployment.
node verify.mjs --online
Where the claim stops
Provenance does not certify scientific truth or independently attest the simulator measurement. A saved snapshot is a past read; --online performs a fresh read. The separate /research-loop journal demonstrates one real GenesisResearch-funded order, delivered callback and simulator experiment. The original review-v2 package remains a historical snapshot.
Ethereum request → · Original IMD artifact → · Contract role and limits →
One file. Fixed bytes.
The immutable archive is the public world state already included as genesis-research-engine/evidence/recorded-world.json in the pinned review ZIP. It contains the two recorded observations and their IMD provenance, without the hidden world opening.
Expected SHA-256:281d59f730460ddf86d075da1e6d39082bb58dad492a2630b9225f31a154d54f
Save both downloads in the same directory. On macOS:
shasum -a 256 -c recorded-world.sha256Expected result: recorded-world.json: OK. The server checks the pinned hash before serving this URL; altered bytes are rejected.
Canonical archive and journal export
The canonical file is byte-identical to recorded-world.json in the pinned review package. The journal export adds format and exportedAt metadata and has a different checksum.
The archive checksum identifies these exact bytes, not who measured them or whether the simulated result is scientifically true.
Download the journal export with metadata → · Read structured evidence levels →