The Taste Gap: Why Senior Engineers Get 10x from AI

Everyone has access to the same AI tools. So why are some engineers getting 10x more out of them?

Ira Glass called it the "taste gap" — the distance between what you can recognize as good and what you can currently produce. He was talking about creative work, but it applies perfectly to software. Right now, it's the single biggest factor in how much value someone gets from AI.

A junior developer and a senior engineer type the same prompt into the same LLM. They get roughly the same output. What happens next is completely different. The senior looks at it and immediately sees the problems: missing error handling, a naive data model, an architecture that won't scale past a hundred users.

The junior ships it.

This isn't a knock on junior developers. They haven't had enough reps to develop the sense of "this doesn't smell right." They don't have the scar tissue from production incidents at 2am, from migrations that went sideways, from that one time the database schema made reporting impossible three months later. That pattern library takes years to build. No amount of AI tooling replaces it.

What AI does replace is the mechanical labor. The senior engineer who used to spend 60% of their time on boilerplate, syntax lookup, and writing the same CRUD operations for the hundredth time? That person now moves ten times faster. Less time typing, more time thinking. They review AI output the way they'd review a junior's pull request — quickly, confidently, with a clear mental model of what "done" looks like.

This matters for hiring and for choosing who builds your software. You don't want the cheapest developer with AI tools. You want the most experienced engineer with AI tools, because they're now doing the work of a small team. AI didn't close the taste gap. It made it wider.