Recorded campaigns compare bounded mutations on historical data. The page exposes their run log; it is not a promise of continuous execution or seven agents running now.
Take the best parameters. Apply small random mutations. Like DNA — most are neutral, rarely one is beneficial.
Replay the mutation on the campaign dataset. Historical jobs used a bounded per-experiment budget; the receipt records what actually ran.
If <code>universal fitness</code> (0-100) improves, keep. Otherwise discard. Only improvements survive.
Each scheduled track preserves its hypothesis, result and rejection reason. No convergence or continuous-autonomy claim is inferred.
The agent that optimizes Darwin itself. Mutation rate, crossover ratio, fitness weights — all auto-tuned. <strong style='color:#ef4444'>Darwin evolves strategies. Meta-Harness evolves Darwin.</strong>
Strategy Arena's Evolution Lab is a read-only archive of research campaigns inspired by autoresearch. Recorded engines such as Darwin, Leviathan, Chimera, Invictus, Hydra, Portfolio and PromptForge mutated parameters and replayed them on historical BTC data. Their outputs are hypotheses and run records, not proof of profitability or a promise of continuous execution. Promotion requires a separate immutable Lab proof, Rust CPU authority and a sealed holdout verdict.