Harness² is a plug-and-play framework from Google Research for recursive agent harnessing in an open world, where agents face tasks unseen during harness design. A harness is the system around a model that lets it act as an agent, including prompts, tools, memory, skills, and orchestration. Harnessing is the process of optimizing that system.
The framework connects two levels of recursion while keeping model weights fixed. Task-level recursion refines the execution harness for the current task through a contrastive Propose–Probe–Compose step: it proposes edits, probes their effects on the same task, and composes a refined harness from behavioral contrasts between executions, without ground-truth feedback. Harnessing-level recursion updates a persistent improvement harness with adopted harnesses and evidence-supported lessons to guide future refinement.
Across three professional benchmarks and four model–harness configurations, a single Harness² step improves GLM-5 and Gemini 3.5 Flash over their base harnesses. Reported gains reach 11.5 points on Work Buddy-Bench, 4.5 points on Job Bench, and 11.8 percentage points on LAB's partial pass rate. Increasing the task-level recursion budget brings further gains.
The release supports parallel and sequential recursion, Open Code and Codex executors, and the Job Bench, LAB, and Work Buddy-Bench benchmarks. Users can preview the framework locally with a demo that requires no model credentials, or run model-backed benchmarks after configuring Vertex AI access.
Source: Hacker News · Summarized by HeadlinesBriefing