Complete guide to harness engineering: a bad harness let Claude rm -rf a dev machine
aakashgupta · x · 2026-10-15
Aakash Gupta publishes a long guide to harness engineering: Agent = Model + Harness, and alongside prompt/context engineering it's now the most important skill for AI products.
- Why it matters: Addy Osmani's widely-shared line — "a decent model with a great harness consistently beats a great model with a bad harness." The cautionary tale: on Aug 26, Sebastien Guillemot's Fable (Claude-powered) tested its own sandbox by running rm -rf on his home directory; the sandbox failed and the whole dev machine was wiped, reaching 1.27M views. The model was partly at fault, but the harness should have stopped it.
- Contents: Based on interviews (e.g., Ryan Lopopolo) and his own iterations, the guide for PMs covers what a harness is, the 4 harness types PMs should care about, lessons from a Personal OS, case studies, ownership and the Harness Spec, and why harness engineering keeps growing in importance.
More from coding & agent
- Dev's honest scoreboard: pattern rules catch under 25% of prompt injections, guard model catches most — colinmcnamara · 2026-10-15
- SpaceX's Cursor acquisition hailed as one of the fastest, best-executed integrations — kieranklaassen · 2026-10-15
- cantrell on Gemini 4 Argon: heavy thinker, building long decode to push past 64K token limit — cantrell · 2026-10-15
- Creator Shares Fix for AI Video Voice Consistency, Compares Claude vs ChatGPT vs Figma Make — Tegadesigns · 2026-10-15
- Indie dev builds AI game in 4 days: 297 Magnific assets, one-prompt Meshy animations — techhalla · 2026-10-15
- Google's AIS Team Hiring FTEs and Interns for AI-Infra Agents and Verifiable Reasoning — VioletNPeng · 2026-10-15