Google's 'Harness²' Lets Weaker Models Beat Stronger Ones by Rewriting Their Own Scaffolding Mid-Task

A new paper shows agents editing their own harness at runtime, with GLM-5 jumping 11.5 points — evidence that better scaffolding can outperform bigger models.

A Google paper on "Harness²," flagged by @itsharmanjot, offers a counterweight to the week's scale-everything narrative. The core finding: weaker models can outperform stronger ones when allowed to dynamically edit their own harness — the surrounding scaffolding of tools, prompts, and control flow — in the middle of a task. GLM-5 reportedly jumped 11.5 points under this approach.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum. Vestibulum ante ipsum primis in faucibus orci luctus et ultrices posuere cubilia curae.

Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. Proin pharetra nonummy pede. Mauris et orci. Aenean nec lorem. In porttitor. Donec laoreet nonummy augue. Suspendisse dui purus, scelerisque at, vulputate vitae, pretium mattis, nunc. Mauris eget neque at sem venenatis eleifend. Ut nonummy. Fusce aliquet pede non pede. Suspendisse dapibus lorem pellentesque magna integer malesuada.

Unlock the full briefing

Get every story in today's briefing, the full archive, and the daily AI intelligence brief.

All stories today

Full archive

Daily brief