ARC-AGI-3 New Strategy: Learn the Rules First
GregKamradt · x · 2026-07-18
This shares a highly practical strategy: when encountering a new task or game, an agent shouldn't rush straight for the goal. Instead, it should spend its initial actions "learning the rules," clarifying all constraints before pursuing the optimal solution.
The context mentions that the [schema] harness achieved very high scores on the ARC-AGI-3 Public set: Opus 4.8 + Fable 5 reached 99% RHAE, and GPT-5.6 Sol hit 95.35%. The poster emphasizes that this methodology of "understanding rules first, then efficiently achieving the goal" might apply not only to ARC-AGI benchmarks but also to real-world tasks.
Related event: Schema Harness Sparks ARC-AGI-3 Debate(14 posts)→
More from Models
- 'Doomer takeover = AI winter': Terra 6.1 flops on Terminal-Bench 4.0 vs Sonnet 5.5 — mark_k · 2026-10-27
- Together AI sits down with Meta to explain Muse Spark Open — togethercompute · 2026-10-27
- GPT-6.1 Terra hits ~60% on Agents' Last Exam at $1/task, 9x cheaper than Opus 5.5 — ChrisGPT · 2026-10-27
- ChatGPT keeps telling tech-support users to email 404 Media's Jason Koebler — petrusenko_max · 2026-10-27
- GPT-6.1 Terra called the awkward middle tier, priced at half of GPT-6.1 Sol — cedric_chee · 2026-10-27
- Mystery model "Terra" scores 48 on Artificial Analysis index with no public API — haider1 · 2026-10-27