Agent Harness Engineering

Addy Osmani argues that the real leverage in AI coding agents lies not in the model itself but in the 'harness' — the scaffolding of prompts, tools, hooks, sandboxes, context policies, and feedback loops wrapped around it. A decent model with a great harness outperforms a great model with a poor one. The post introduces 'harness engineering' as a discipline: treating every agent mistake as a permanent signal that tightens the harness (the ratchet principle), designing components by working backwards from desired behaviour, and maintaining AGENTS.md as a concise, failure-traced rulebook. Key harness primitives covered include filesystem/Git for durable state, bash execution, sandboxes, memory/search, context compaction strategies, hooks for enforcement, planner/evaluator splits, and the emerging Harness-as-a-Service model. The post concludes that as models improve, harness complexity doesn't shrink — it moves to address newly unlocked capabilities.