Evidence

One number, one mode, one ledger.

Strict, bounded, frontier, campaign, atlas, and historical evidence answer different questions. This page refuses to blend them.

Strict first move

1108.8 Elo

Reported result; one frozen first executable artifact, before task-time search.

Sealed blind evaluation ledger

Strict field rank

14/98

Reported result in the sealed field.

Sealed blind evaluation ledger

Bounded probe

1212.2 Elo

Reported result; bounded task-time selection, not a strict first move.

Sealed bounded-probe ledger

Bounded field rank

7/98

Reported result in the sealed field.

Sealed bounded-probe ledger

Frontier diagnostic

1331.1 Elo

Diagnostic upper frontier; not a deployable first-move result.

Sealed frontier diagnostic ledger

Frontier field rank

6/98

Diagnostic rank, kept separate from strict and bounded modes.

Sealed frontier diagnostic ledger

Outcome-ranked atlas

600 cases

353 wins and 247 losses; deliberately outcome-ranked and non-representative.

Outcome-ranked atlas ledger

Full matched cohort

1,649 cases

353 wins and 1,296 losses in the complete matched historical record.

Full matched-cohort build manifest

Training campaign

36 generations

Company-reported campaign record.

Self-play campaign ledger

Validated promotions

7 promotions

Company-reported retained-frontier promotions.

Self-play campaign ledger

Processed rows

375,522 rows

Summed processed rows across the recorded campaign.

Self-play campaign ledger

Generation-hours

36.42 hours

Summed generation-hours; not critical-path wall time or deployment time.

Campaign time derivation ledger

Selection boundaries

The outcome-ranked atlas contains 600 deliberately selected cases: 353 wins and 247 losses. It is useful for inspecting mechanisms and failures, but it is explicitly non-representative. The full matched historical cohort contains 1,649 cases: 353 wins and 1,296 losses.

Published data

Evidence ledger v1 (JSON) · Import provenance v1 (JSON)

Imported artifact custody

ArtifactSHA-256Bytes
Move42_InnovationZero_v1.pdf7a80bbfac4f20a9db7b94bd46111fa6262d72bc65dbcf33c3605a484f68dabce605448
sealed_blind_ledger.jsonb427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e193838281761
campaign_ledger.json7f34b6f6b695bc70e068b7bbb4d7ed563ebf22c12edb7c7983eb05c774e3349b9415
build_manifest.jsone5e5f695cce8ebd39467bfc1819838c2b0433ec3f37ac190ccef9f618a59175b577
claim_ledger.jsond55133c561b531681aeeeef4c2b35e5653c9b6dd56cf02b82bf8bd9ad480d45127839
atlas_analysis.json06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120523777
claim_build_receipt.json6943113a36d794c906b3b154264a0ec06e023597ba909fb40ea6c0f66e8765e8431
campaign_claims.jsondebd62e215a0bf05a1b1028f19838d9dbbb19804d77450471f15135be485df5c83797
source_registry.json9fab13ee654d73117d2297d2d8b70c1912e2011765c8afa44821caf2fa3f8c809217
citation_ledger.jsonbb7a6fac22c38b0c10b3ffe488d4c907af77c79046f7384f8b52264a90fb3b0a3640

Limits

These are company-reported results and a historical record. They do not establish causal mechanisms, universal superiority, safe autonomous use, or performance on untested distributions. No corpus-scale candidate-count claim is published because no public corpus-scale receipt is present.