harness engineering
30 articles about harness engineering in AI news
Agentic Harness Engineering Boosts Coding Agents 7% on Terminal-Bench 2
Agentic Harness Engineering introduces a structured approach to evolving coding-agent harnesses, using revertible components, condensed experience, and falsifiable decisions. On Terminal-Bench 2, pass@1 climbs from 69.7% to 77.0% in ten iterations, beating human-designed baselines.
Harness Engineering for AI Agents: Building Production-Ready Systems That Don’t Break
A technical guide on 'Harness Engineering'—a systematic approach to building reliable, production-ready AI agents that move beyond impressive demos. This addresses the critical industry gap where most agent pilots fail to reach deployment.
Agent Harness Engineering: The 'OS' That Makes LLMs Useful
A clear analogy frames raw LLMs as CPUs needing an operating system. The agent harness—managing tools, memory, and execution—is what creates useful applications, as proven by LangChain's benchmark jump.
Meta-Harness Framework Automates AI Agent Engineering, Achieves 6x Performance Gap on Same Model
A new framework called Meta-Harness automates the optimization of AI agent harnesses—the system prompts, tools, and logic that wrap a model. By analyzing raw failure logs at scale, it improved text classification by 7.7 points while using 4x fewer tokens, demonstrating that harness engineering is a major leverage point as model capabilities converge.
Open-Source Course Shows Harness, Not Model, Lifts Coding Agent 25 Places
Open-source course shows harness engineering, not model swap, moved a coding agent from ~30th to top 5 on Terminal-Bench. Course builds Decode from scratch.
Anthropic Deploys Multi-Agent Harness to Scale Claude's Frontend Design & Autonomous Software Engineering
Anthropic engineers detail a multi-agent system that orchestrates multiple Claude instances to tackle complex, long-running software tasks like frontend design. The approach aims to overcome single-model context and reasoning limits.
Claude Code Reverse-Engineered: 98.4% of Codebase is Operational Harness
A reverse-engineering analysis of Claude Code reveals only 1.6% of its codebase is AI decision logic, with the rest being operational infrastructure. This challenges current agent design paradigms by prioritizing a robust deterministic harness over complex model routing.
Agent Harness Debate: Anthropic vs. OpenAI vs. LangChain on Scaffolding
A central debate in agent engineering pits a 'thin harness' approach (Anthropic) against 'thick harness' designs (LangGraph). The infrastructure layer, not the model, is becoming the primary product differentiator.
5 Harness Internals That Changed How I Use Claude Code Daily
Rebuilding Claude Code's harness reveals that CLAUDE.md layers on a hidden base prompt, hooks can block tool calls, and subagents need abort trees—5 actionable takeaways for daily use.
Stanford, Meta 'Code as Agent Harness' Paper Rethinks AI Agent Design
Stanford and Meta's "Code as Agent Harness" paper proposes code-driven AI agent orchestration, potentially improving reliability over natural language prompts.
Meta-Stanford Survey: Code as Agent Harness Improves AI Reasoning
Meta, Stanford, Illinois survey argues AI agents work better with code as their main working layer, calling it an agent harness.
Claude Code's Six-Layer Architecture: Harness, Not Magic
Claude Code's six-layer architecture uses a 3-layer context compressor at 92% threshold and Redis-based multi-agent FSM protocol. The model is just one node in a harness.
Agent Harnessing: The Infrastructure That Makes AI Agents Work
A detailed technical guide argues that the model is not the hard part of building AI agents. The six-component harness — context management, memory, tools, control flow, verification, and coordination — is what separates production-grade agents from those that fail silently.
Akshay Pachaar Inverts LLM Agent Architecture with 'Harness' Design
AI engineer Akshay Pachaar outlined a novel 'harness' architecture for LLM agents that externalizes intelligence into memory, skills, and protocols. He is building a minimal, didactic open-source implementation of this design.
Stanford/MIT Paper: AI Performance Depends on 'Model Harnesses'
A new paper from Stanford and MIT introduces the concept of 'Model Harnesses,' arguing that the wrapper of prompts, tools, and infrastructure around a base model is a primary determinant of real-world AI performance.
Meta-Harness from Stanford/MIT Shows System Code Creates 6x AI Performance Gap
Stanford and MIT researchers show AI performance depends as much on the surrounding system code (the 'harness') as the model itself. Their Meta-Harness framework automatically improves this code, yielding significant gains in reasoning and classification tasks.
MiniMax M2.7 AI Agent Rewrites Its Own Harness, Achieving 9 Gold Medals on MLE Bench Lite Without Retraining
MiniMax's M2.7 agent autonomously rewrites its own operational harness—skills, memory, and workflow rules—through a self-optimization loop. After 100+ internal rounds, it earned 9 gold medals on OpenAI's MLE Bench Lite without weight updates.
Anthropic's 'Harness' Design: What It Means for Your Claude Code Workflows
Anthropic's new 'harness' architecture for long-running apps could enable Claude Code to manage more complex, persistent development tasks with greater stability.
Google DeepMind's AutoHarness: The AI Tool That Could Revolutionize How We Build Intelligent Systems
Google DeepMind's AutoHarness framework enables automatic testing and optimization of AI models without retraining, allowing developers to synthesize functional AI agents like coding assistants with unprecedented efficiency.
The AI Paradox: Why Software Engineering Jobs Are Surging Despite Automation Fears
Citadel Securities data reveals software engineering job postings are spiking despite AI coding tools, illustrating the Jevons paradox where cheaper software creation drives increased demand for developers as companies expand digital initiatives.
Agent Harness Scaling: EFC Predicts Success at R2 0.99 vs 0.42
New research introduces Effective Feedback Compute (EFC), which predicts agent success at R2 0.99 vs 0.42 for raw tokens. Reallocating compute by EFC lifts success 3x at the same budget.
How to Manage Multiple Claude Code Sessions with Harness and Preview
Two actionable tools to solve the core productivity bottlenecks when running multiple Claude Code agents: session management and review speed.
The AI Paradox: How Cheaper Code Creation Is Fueling a Software Engineering Boom
Contrary to fears of AI replacing developers, the Jevons Paradox suggests that making software creation cheaper through AI tools actually increases demand for human engineers who can design, review, and integrate complex systems at scale.
Databricks Tests Coding Agents on Its Own Codebase
Databricks benchmarked coding agents on its own polyglot codebase. GLM-5.2 matched top closed models, a minimal harness halved costs, and cheaper-per-token models cost more per task.
Stop Testing Skills Once: Use Caliper's pass@k to Measure What Actually
Caliper is a lightweight harness that runs Claude Code skills k times, scores them with pass@k, and compares against a no-skill baseline so you know if your skill actually helps.
How Claude Code's System Prompt Engine Actually Works
Claude Code builds its system prompt dynamically from core instructions, conditional tool definitions, user files, and managed conversation history, revealing the critical role of context engineering.
Beijing Unveils 10-Measure Agent AI Policy With Token Economy
Beijing's July 2026 Agent AI policy introduces 10 measures including token economy infrastructure, signaling regulated economic framework for autonomous AI systems.
Build a Self-Sustaining Claude Code Environment: The Complete 14-Part System
Build a self-sustaining Claude Code environment with 14 components: memory, skills, autonomy, guardrails, and monitoring. Connect them into a feedback loop where measurements flow back into memory. Use CLAUDE.md and hooks.
Harbor Adds LangSmith Sandbox Support, Making Agent Eval Backends Swappable
Harbor, an open-source agent-evaluation framework, now integrates LangSmith sandboxes. This allows users to run the same eval across multiple providers (Daytona, Modal, E2B, LangSmith) with a single flag change, eliminating per-provider setup tax.
MirrorCode Rebuilds Programs from Behavior Alone, Beats GPT-4o by 37%
Epoch AI's MirrorCode reconstructs programs from I/O behavior alone, scoring 67.3% on SWE-bench—37% above GPT-4o—without source code or traces.