• Harness Engineering Birgitta is a Distinguished Engineer and AI-assisted delivery expert at Thoughtworks. • She has over 20 years of experience as a software developer, architect and technical leader. • This article is part ofâ Exploring Gen AIâ . • A series capturing Thoughtworks technologists’ explorations of using gen ai technology for software development. • 17 February 2026 It was very interesting to readOpenAIâ s recent write-up on â Harness engineeringâ which describes how a team used â no manually typed code at allâ as a forcing function to build a harness for maintaining a large application with AI agents. • After 5 months, theyâ ve built a real product thatâ s now over 1 million lines of code.
Article Summaries:
- OpenAI’s recent “Harness Engineering” write‑up details how a team built a million‑line‑code product using only AI agents, with no manual coding. The approach relies on a “harness” that blends deterministic tools and LLM‑based agents across three layers: context engineering (dynamic knowledge bases and observability), architectural constraints (custom linters and structural tests), and garbage‑collection agents that spot documentation or constraint violations. The process is iterative-when an agent fails, missing guardrails or documentation are added and the fix is generated by Codex itself. The author suggests harnesses could evolve into reusable service templates, raising questions about sharing and synchronizing updates across teams.
Sources: