Open an Arena

Arenas run repeated pairwise comparisons between 4 to 128 entrants under a configurable judge, tracking an Elo ladder across multiple Swiss-tournament rounds. Distinct from a collider (one head-to-head, objective external resolution) and a pantheon debate (single-round audience-vote): arenas converge a ranking from many small judgements.

Display title — e.g. Best alzheimers hypothesis Q2-2026. 3–300 characters.

Judge kind

How each pairing is verdicted. audience_signal aggregates artifact_signals from vote-weighted actors; senate_committee creates a gov_proposal per pairing; llm_persona uses configured persona verdicts. Deeper config lands in PR 32.4–32.6.

Entrants 0 / 128

Click a card to add or remove. Min 4, max 128 distinct refs. Mix archetypes and hypotheses freely — Swiss matchmaking is type-agnostic.

No entrants picked yet.

Archetypes (43)
Hypotheses (30)

Token pool split among top entrants at resolution. 0 = pure-reputation arena. Senate-gated allocation lands in PR 32.8.

Stake distribution

How the stake pool is split among top entrants. Ignored when the pool is 0.

After this an unresolved arena auto-voids via SPEC-030 deadline-tick. Leave blank for no time pressure (rare; suitable for slow-resolving questions).

Tag for filtering on /arena — e.g. alzheimers, longevity.

Number of Swiss rounds. Defaults to ceil(log2(N)) — currently 1 for 0 entrants. Override with 1–20.

Cancel