Tech lead lays out an AI-assisted workflow that forces thinking before typing

The problem

It has been reported that a Tech Lead shared a structured AI-assisted development workflow on Hacker News and his personal blog after growing tired of the familiar trade-off: speed at the cost of understanding. The scenario is maddeningly common — you open a chat, iterate fast, ship something that technically works, and then realize nobody (including you) really knows what lives in production. Edge cases go unhandled, architecture crumbles with the next feature. Move fast and break things, indeed. The emotional punch here is simple: building faster shouldn’t mean understanding less.

The workflow

The core idea is blunt and elegant — think in writing, not in code. The author says AI is great at implementation but lousy at figuring out what you actually want; that part is and remains human work. The process starts with a free-form plan written in plain language, then runs that through a "write-a-prd" skill that allegedly inspects the codebase and conducts a meticulous interview to turn hand-waving into explicit decisions. That PRD produces user stories, implementation decisions, testing plans, and explicit out-of-scope items. Finally, a "prd-to-issues" step chops the PRD into vertical slices — end-to-end, demoable issues — and tags each as AFK (AI-first) or HITL (human-in-the-loop).

Why it matters

This isn’t just another AI productivity hack; it’s a countermeasure to cargo-cult coding. By forcing the thinking upstream and using AI to stress-test decisions rather than replace them, the workflow aims to preserve intentionality while reclaiming speed. It has been reported that some teams are experimenting with similar patterns; the real question now is cultural: will organizations treat AI as a power tool or hand it the wheel? Either way, this roadmap offers a practical middle ground — faster shipping, but with paperwork that actually explains what you shipped.

Sources: maiobarbero.dev, Hacker News