Yoav Goldberg: Ark's harness was simply bad, and OpenAI's fix was obvious
yoavgo · x · 2026-09-05
Researcher Yoav Goldberg (yoavgo) weighed in on a recent model controversy, arguing that the claimed importance of the harness turned out to be largely nonsense — Ark's harness was simply bad — and OpenAI's "fix" was the obvious response.
- He adds that capabilities that previously depended on the harness now appear to have been pushed into the model itself
- A notable first-hand researcher take on a live models story
More from Models
- Official code repo for "Hands-On Large Language Models" hits 28.9k stars — ZabihullahAtal · 2026-09-05
- Dev mocks ARC-AGI's escalating claims: from 3-year-old AGI to hypothetical StarCraft AGI — AndrewDai · 2026-09-05
- Claude completes first formalized proof of Fermat's Last Theorem in 13M lines of Lean — AnthropicAI · 2026-09-05
- Fable 5.1, GLM-5.3 open weights, and three real-time world models headline a packed AI week — thursdai_pod · 2026-09-05
- Greg Brockman closes GPT-6 briefing with "welcome to the AGI era" — thursdai_pod · 2026-09-05
- GPT-6 computer use demoed live: macOS simulator "it works, it's real" — thursdai_pod · 2026-09-05