Agent Harness Aces ARC-AGI-3
naval · x · 2026-07-16
The Impossible Research team introduced an agent harness capable of playing games, writing code, and reasoning "like a physicist."
The post cites [schema] scores: achieving 99% RHAE on the ARC-AGI-3 Public set using Opus 4.8 + Fable 5, and 95.35% using GPT-5.6 Sol. The author's central claim is that this harness organizes LLMs into workflows better suited for high-difficulty reasoning and task execution.
Related event: Schema Harness Sparks ARC-AGI-3 Debate(14 posts)→
More from coding & agent
- Anthropic researcher: 99% of engineers now run swarms of 300+ self-improving agents — AlishaOutridge · 2026-09-11
- Gergely Orosz: Shipping 10x PRs With AI Agents, Sites Fill With Small Regressions — ducha_aiki · 2026-09-11
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Astra storyboards plus Minimax H3 per-shot generation boost video success rates — Hailuo_AI · 2026-09-11
- Codex tip: use Sol with Astra and Luna sub-agents to save usage — pvncher · 2026-09-11
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11