Public results / evidence-state index
This hub points to the public leaderboard, figure package, and terminal experiment outcomes. It keeps frozen, terminal-data, provisional, and design states separate so an available artifact never becomes a stronger claim by navigation alone.
Evidence state key
States describe what the public artifact authorizes a reader to say. They are labels with different jobs, not a single progress bar.
The public artifact is frozen for the claim it names.
The data are terminal; analysis qualifications stay attached.
A dated readout is visible, but it is not a terminal claim.
The instrument or preflight exists before model outcomes.
Public evidence surfaces
The public cohort table carries its opponent, color, clock, clean-game rule, and cohort state with every estimate.
Boundary. These are local protocol-specific measurements, not FIDE, online-platform, or general model ratings.
Open the leaderboardThe gallery preserves the terminal-frozen Luna figure alongside its caption, accessibility text, and artifact identity.
Boundary. Terminal frozen, exploratory, and protocol-specific.
Open the Luna terminal figureThe public figure package preserves a historical interim and unfrozen DeepSeek projection. Terminal data, a rating, and the joint E000/E001 qualification exist. The public DeepSeek cohort remains interim; public Elo and ranking are withheld, publication is unauthorized, and request-level billing attribution is unavailable.
Boundary. Historical interim, unfrozen, and provisional snapshots from the pre-terminal synthesis; terminal DeepSeek data and publication authorization are represented separately.
Inspect the DeepSeek interim figureDoes redundant access to the root board or complete legal-move list change system regret when the underlying chess state, action universe, response contract, and request length stay fixed?
Boundary. R3's installed Responses path remains the observed NO-GO; R3A is isolated and not integrated into the live harness. No provider-received bytes, provider compatibility, provider behavior, model effect, chess-quality effect, power result, final-position result, paid execution, or causal benefit has been observed.
Explore the design preflightPublic terminal artifacts / terminal aggregates
E025
Across 48 positions and 192 registered calls, changing candidate order produced 42.7% disagreement; identical-order baselines were 39.6% and 35.4%.
The terminal aggregate is not clearly supported as a non-zero effect.
Claim boundary. Action stability for one dated Luna alias cohort on development-exposed Black positions; not move quality, Elo, search, or a universal null.
E028
Across 48 paired positions, visible fixed-set membership was 95.8% versus 2.1% in the board-only arm.
Membership in the supplied set, not move quality or chess strength.
Claim boundary. Development-only causal effect of showing one fixed five-move suggestion set on Luna set adherence; no move-quality, Elo, vision, search, or candidate-generation claim.
E029
Across 60 complete paired positions, exact-top-two advice changed mean 50-move-rule utility from 42.5% to 49.6%.
positive confirmatory; the analyzer qualification remains public.
Claim boundary. A local causal effect of displaying two unranked exact-policy moves on exact 50-move-rule move utility for this cohort; no full-game transfer, Elo, learning, vision, search, or general chess-strength claim.
A113
Across 319 complete pairs from 320 source puzzles and 640 scheduled episodes, the fixed-budget Stockfish scaffold changed complete-episode survival by +11.6 pp for the same Luna model.
Model-plus-harness protocol effect; not model-only learning or Elo.
Claim boundary. A positive paired engine-scaffold effect on registered puzzle-decision survival for this Luna cohort; no learning, Elo, unassisted-strength, or general agentic-chess claim.