Writing

We treat AI like a vending machine. Prompt in, code out. When it fails, we prompt harder. That's nonsensical.

I built a dashboard to track PR review times. What I found broke my assumptions about code review culture. Large PRs were getting approved almost as fast as small ones ... and that says everything about where we're actually paying attention.

Accountability means knowing why. When AI breaks something, there's no 'why' to interrogate. Until you define it.

Every productivity system I tried assumed my brain could be disciplined into submission. They were all solving the wrong problem.

Every junior engineer thinks the job is syntax. Every senior engineer knows it's the conversation three weeks before the syntax. AI just made that clearer than ever.

The strongest engineers walking through the door are not the ones with the cleanest GitHub. They're the ones who can describe, in detail, the agent loops they've built.

The loudest voices in engineering keep framing AI as the great equalizer that will let mediocre developers ship junk code. They're missing what's actually happening.

Six months into AI adoption, every team has dashboard metrics that look good. But ask what a good AI-assisted PR looks like and you'll get six different answers. The tools showed up, people figured out their own way, and no one had the conversation about what 'good' actually means.

AI changes speed. Not judgment. If a team already struggled to make sound architectural decisions, the tool doesn't rescue them. It just helps them make more bad decisions faster.