AI and Agentic Coding

AI made code generation cheap. It did not make understanding cheap. These episodes work through what actually changed.

The pattern across these conversations: AI absorbs a lot of the pain that made legacy work miserable. It can read the commits and the old tickets and explain why a system behaves the way it does, then write tests to lock that behavior in. What it does not touch is the part that actually sinks rewrites: changing code nobody fully understands. Vague requirements still produce vague systems, only faster.

Recent episodes get into agentic workflows, what changes when acceptance criteria are written for an agent instead of a person, and where the security exposure sits. The guide covers whether AI changes the rewrite math.

30 episodes · browse the full archive

Other topics