Harness Engineering

Harness Engineering could be a popular term shortly. I've just read an interesting post about OpenAI building and shipping an internal product with zero human-written code. Every line was generated by agents, with the developers focused on intent, guardrails and feedback loops for the AI.

I'm not sold on fully AI-built enterprise systems today. But it's hard to ignore where things are heading. Even if LLM capability plateaued tomorrow, it feels like we're at the stage where it's more about assembling the right scaffolding around them.

The real opportunity might be in building the services, tooling and guardrails that let AI validate its own work across security, performance, extensibility, accessibility and consistency. Less about replacing engineers, more about engineering the environment the AI operates in.

I can easily imagine a future where you ask for an enterprise feature and get five solid, production-ready options back to evaluate. Your focus shifts toward product thinking and UX, not the mechanics of the code itself.

The article is linked in the comments of the original LinkedIn post.