# The Last Harness You'll Ever Build

arXiv: https://arxiv.org/abs/2604.21003 · HTML: https://mlsanity.com/paper/2604.21003/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 2 of its 10 references that are in the map.
- First posted on arXiv 22 Apr 2026.

## Builds on (most cited first, 2 of 2)

- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [OpAgent: Operator Agent for Web Navigation](https://mlsanity.com/paper/2602.13559/) (2026)

## Built on by (most cited first, 6 of 6)

- [Learning to Control LLM Agent Harnesses with Offline Reinforcement Learning](https://mlsanity.com/paper/2607.05458/) (2026)
- [VERSE: Verified Self-Evolving Optimizer for Agent Harnesses](https://mlsanity.com/paper/2610.02616/) (2026)
- [ToFu: A White-Box, Token-Efficient Agent Harness for Researchers](https://mlsanity.com/paper/2607.11423/) (2026)
- [Compile, Don't Memorize: A Context Compilation Architecture (CCA) for In-Context Learning](https://mlsanity.com/paper/2609.00759/) (2026)
- [Harness Learning Enables Generalizable Test-Time Adaptation](https://mlsanity.com/paper/2609.35738/) (2026)
- [Tracing the Thoughts of a Coding Agent Playing ARC-AGI-3: Lessons for Continual Learning](https://mlsanity.com/paper/2610.11450/) (2026)

## Abstract (arXiv)

AI agents are increasingly deployed on complex, domain-specific workflows -- navigating enterprise web applications that require dozens of clicks and form fills, orchestrating multi-step research pipelines that span search, extraction, and synthesis, automating code review across unfamiliar repositories, and handling customer escalations that demand nuanced domain knowledge. Each new task domain requires painstaking, expert-driven harness engineering: designing the prompts, tools, orchestration logic, and evaluation criteria that make a foundation model effective. We present a two-level framework that automates this process. At the first level, the Harness Evolution Loop optimizes a worker agent's harness H for a single task: a Worker Agent WH executes the task, an Evaluator Agent V adversarially diagnoses failures and scores performance, and an Evolution Agent E modifies the harness based on the full history of prior attempts. At the second level, the Meta-Evolution Loop optimizes the evolution blueprint Lambda = (WH, H(0), V, E) itself across diverse tasks, textbflearning a blueprint Lambda(best) that enables rapid harness convergence on any new task -- so that adapting an agent to a novel domain requires no human harness engineering at all. We formalize the correspondence to meta-learning and present both algorithms. The framework shifts manual harness engineering into automated harness engineering, and takes one step further -- automating the design of the automation itself.