An Engineer's Reflection: How AI Agents Quadrupled His Prototyping Velocity

A working engineer documents how AI coding agents have reshaped his personal workflow over the past year, estimating a roughly 4x speedup on time-to-PR for typical tasks. The more interesting shift, he argues, isn’t raw speed but scope: ideas that would have been parked as ‘nice but no time’ now fit into an afternoon, and his GitHub has filled with running prototypes spanning a systems language, a config notation, a secrets CLI, and a block editor. The same velocity has carried into his day job, where he credits agents with letting him ship engineer-support automation and cut internal codespace bootstrap times in half.

The author frames the deeper change as a move up the abstraction ladder. Not typing every line forces him to think in contracts, boundaries, and written specs, which turns out to be the same muscle as delegating to a junior engineer. He’s wary of the obvious trap: if the tools do everything, his own technical dexterity atrophies, so he deliberately carves out time to read source, sit with a debugger, and implement things by hand.

He stays cautious about the environmental, financial, and social questions around AI, and avoids any tool pitch. The piece nods to similar reflections from Mike McQuaid, Cassidy Williams, and Simon Willison, positioning itself as one data point in a broader shift in how individual engineers are choosing to work.