Tycho Coding-Agent Achieves 88.49 RHAE on ARC-AGI-3
July 31, 2026
Tycho is a coding-agent system that constructs programmatic world models to solve ARC-AGI-3 tasks by modeling environments as deterministic Moore machines. Using Claude Opus 4.8, the agent achieved a Relative Human Action Efficiency of 88.49 through actor-requested model delegation.
HOW THIS AFFECTS YOU
●
builderYou can leverage agentic orchestration to build systems that learn and adapt to new rules through code generation.
●
researcherThis demonstrates the efficacy of using executable hypotheses for reasoning in complex, rule-based environments.