---
title: "Google open-sources Harness2, which rewrites agent harnesses at test time"
url: https://www.parallelquant.com/posts/google-open-sources-harness2-which-rewrites-agent-harnesses-at-test-time-1ab254
source_name: "MarkTechPost"
source_url: https://www.marktechpost.com/2026/10/18/google-harness2-recursive-agent-harnessing-framework/
published: 2026-10-18T08:15:04.000Z
topics: ["open source", "research"]
publisher: "Parallel Quant"
---

# Google open-sources Harness2, which rewrites agent harnesses at test time

*2026-10-18 · Source: [MarkTechPost](https://www.marktechpost.com/2026/10/18/google-harness2-recursive-agent-harnessing-framework/)*

Google Research released Harness2 under Apache-2.0, a framework that rewrites an agent's harness at test time. One refinement step lifted GLM-5 by 11.5 points on WorkBuddy-Bench with no weight updates.

**Why it matters:** Gains from improving the scaffolding around a model, not the model itself, are cheap and usable with any open weights. It reinforces that harness design is a major performance lever for agents.

**Topics:** open source, research

---
Read the original: https://www.marktechpost.com/2026/10/18/google-harness2-recursive-agent-harnessing-framework/
Canonical: https://www.parallelquant.com/posts/google-open-sources-harness2-which-rewrites-agent-harnesses-at-test-time-1ab254
Published by Parallel Quant — https://www.parallelquant.com