🤖Harness²: Recursive Agent Harnessing Framework Released
Recursive agent harnessing just got a major upgrade
TL;DR
Harness² is a plug-and-play framework for recursive agent harnessing, improving GLM-5 and Gemini 3.5 Flash by up to 11.5 points. Supports parallel and sequential recursion, OpenCode and Codex executors, and JobBench, LAB, and WorkBuddy-Bench.
Harness², a plug-and-play framework for recursive agent harnessing, is now available. It connects two levels of recursion while keeping model weights fixed, improving GLM-5 and Gemini 3.5 Flash by up to 11.5 points on WorkBuddy-Bench. This framework is a game changer for developers working with large language models, as it allows for dynamic refinement of execution harnesses and persistent improvement harnesses. The framework supports parallel and sequential recursion, OpenCode and Codex executors, and benchmarks like JobBench, LAB, and WorkBuddy-Bench. With Harness², developers can optimize their models' performance without altering the underlying model weights.
Key Points
Harness² connects two levels of recursion while keeping model weights fixed, improving GLM-5 and Gemini 3.5 Flash by up to 11.5 points on WorkBuddy-Bench.
The framework supports parallel and sequential recursion, OpenCode and Codex executors, and benchmarks like JobBench, LAB, and WorkBuddy-Bench.
Harness² exposes eight editable components around the model: system prompt, guardrails, memory, tool guidance, subagents, skills, scripts, and plugins.
The local demo runs both recursion modes with deterministic editors and real Python executions, requiring no model credentials.
For model-backed benchmarks, install Node.js 20+, the Google Cloud CLI, and benchmark dependencies; LAB needs Pandoc and Poppler; WorkBuddy-Bench needs Docker.
Why It Matters
If you're working with large language models like GLM-5 or Gemini 3.5 Flash, Harness² can significantly improve performance by up to 11.5 points on WorkBuddy-Bench. This framework allows for dynamic refinement of execution harnesses and persistent improvement harnesses, making it a must-have for developers looking to optimize their models' performance without altering the underlying model weights.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,567 builders reading daily.