Google Researchers Open-Source Harness² to Improve Agent Harnesses During Execution

1 reports 1 reporting sources 13 hours agoNew

Understand the event

AI overview

Google researchers introduced and open-sourced Harness², a system that modifies agent harness code, system prompts, tools, and skills during execution; reported WorkBuddy-Bench results show GLM-5 with OpenCode improving from 64.2 to 75.7 after a single improvement step. Harness² supports OpenCode and Codex. It proposes, probes, and combines changes for individual tasks, while maintaining a persistent library of experience from improvements. In the reported benchmark comparison, the improved GLM-5 setup exceeded the unmodified Gemini 3.5 Flash score of 72.4.

Generated by AI from reports · Updated 52 minutes ago

Latest development10/19 11:11 (UTC+8)

Google open-sources Harness² research on recursive agent harness improvement 量子位

Models in this story

Gemini 3.5 Flash$1.5 input / $9 output per million tokens

API prices of all models

Report timeline

Follow the reports to explore different sides of the event.

10/19

11:11
量子位Picks
Google open-sources Harness² research on recursive agent harness improvement
Google researchers introduced and open-sourced Harness², a system that modifies agent harness code, system prompts, tools, and skills during execution. It combines task-level proposal, probing, and composition with a persistent library of improvement experience, supporting OpenCode and Codex. The reported WorkBuddy-Bench results show GLM-5 improving from 64.2 to 75.7 with OpenCode after a single improvement step, exceeding the unmodified Gemini 3.5 Flash score of 72.4.

The same kind of briefing as the one above: a few stories worth reading. Free. Mail starts after you confirm, and every letter has a one-click unsubscribe. Read a past issue

Event heat trend

There is not enough continuous observation data to plot a trend yet.

Keep reading

Latest daily · Issue 26VAST open-sources EdenLive world model with over 24FPS generation at 720P