June 6, 2026
Ctrl+Alt+DRAMA
Harness engineering: Leveraging Codex in an agent-first world
OpenAI says its bots built the app fast — commenters say “absolutely not”
TLDR: OpenAI says three engineers used Codex to build and ship an app with no hand-written code, reaching about a million lines in months. Commenters are split between awe and alarm, with critics calling it bloated, risky, and hilariously under-edited — including the blog post itself.
OpenAI just dropped a brag post saying a tiny team built an internal app with zero hand-written code, letting Codex do everything from the first setup to tests, tools, and docs. The company says the result grew to about a million lines in just months and shipped in roughly one-tenth the usual time. In plain English: humans gave directions, the AI did the typing. But in the comments, the real show began — and readers were not quietly impressed.
The loudest reaction was pure horror. One commenter flatly declared, “I would never dare put that in production,” basically summing up the trust issue in one sentence. Others went for mockery, asking whether it really takes a million lines to make what one person called a “half baked crud app,” a jab at a basic create-read-update-delete business tool. And then came the literary criticism: one reader sniped that if OpenAI wrote zero lines of code by hand, maybe it also wrote zero lines of the blog post by hand, roasting the article itself as an endless parade of AI buzzwords.
Still, not everyone joined the pile-on. One more sympathetic voice argued that critics were missing the point, praising the project’s instruction files and quality checks as potentially better than the usual chaos many teams live with. So the drama is set: is this the future of building software, or just a very expensive way to mass-produce spaghetti? The comments section on OpenAI’s post chose violence.
Key Points
- •OpenAI says it built and shipped an internal beta software product over five months with zero lines of manually written code.
- •The product has internal daily users and external alpha testers, and all code for application logic, tests, CI, documentation, observability, and tooling was generated by Codex.
- •OpenAI estimates the product was built in about one-tenth the time it would have taken to write the code manually.
- •The article says engineers’ role shifted from writing code to designing environments, specifying intent, and building feedback loops for reliable agent execution.
- •The first commit landed in late August 2025, and the initial repository scaffold was generated by Codex CLI using GPT-5 and existing templates.