arXiv: Harness^2 — recursive agent harnessing for open-world tasks
rohanpaul_ai · x · 2026-11-15
Rohan Paul links the arXiv paper "Harness^2: Recursive Agent Harnessing for an Open World" (2610.14224) by Northwestern and Google Cloud AI Research researchers. Same paper as covered in detail in a parallel post: a plug-and-play framework with task-level and harnessing-level recursion that self-improves agent harnesses without ground-truth verifiers, yielding up to +11.5 points on WorkBuddy-Bench.
More from coding & agent
- Proof of Agent goes viral: paste a prompt and let your AI agent mine POA — adamamcbride · 2026-11-15
- Google and Northwestern's Harness² self-improves agent harnesses without answer keys — rohanpaul_ai · 2026-11-15
- Dev builds a PowerShell AI harness for Microsoft's new Intelligent Terminal — dfinke · 2026-11-15
- snitch: a harness that watches agents in parallel and reverts bad paths — generativist · 2026-11-15
- Dev: unattended agent runs burn 99% of tokens on useless reproducibility gates — generativist · 2026-11-15
- workspace-cli: open-source Rust CLI for AI agents to drive Google Workspace — majidmanzarpour · 2026-11-15