A clearer view of AI.
THE INDEPENDENT AI.

COMMUNITY PICK · 19 OCTOBER 2026

@Saboo_Shubham_: Google Research open-sources Harness2

19 October 2026, 11:58 · 2 sources · Curated by Sara Simeone

Google Cloud AI Research released Harness2, a free, open-source framework that automatically improves the setup around an AI agent (its instructions, guardrails, memory, tools, skills and helper agents) without retraining the model. It tries changes on a task, compares the results and keeps lessons for next time. The researchers report gains of up to 11.5 points on work-style benchmarks; those are their own results, and it's research code, not a supported Google product.

Why it mattersIt's the direction agents are heading: the setup around the model matters as much as the model. Technical teams can learn from it; most businesses should take the idea, not the code.

Curated by Sara Simeone as part of our human editorial coverage. The original sources are linked below; curation is distinct from a Hype or Fact? assessment.

Sources

Related AI tools

How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.

MORE FROM 19 OCTOBER 2026

  1. FROM THE COMMUNITY@kimmonismus: China drafts export rules for its biggest AI models
  2. FROM THE COMMUNITY@kimmonismus and @Mr_Salio: three unreleased OpenAI models spotted
  3. FROM THE COMMUNITY@mattpocockuk: a one-prompt guardrail for AI-written code
  4. FROM THE COMMUNITY@ctatedev: agent-browser now sends only what changed on a page
All community picks ↗