On AI

I've spent twenty years building systems where the stakes were physical: dispatch, surgery, pet health, financial settlement. In every one of those domains, the question "what happens when the system is wrong?" had an answer measured in something other than money.

AI is a tool that has changed how I lead engineering teams. It hasn't changed what good engineering is. The systems I build still have to fail safely, leave an audit trail, and survive the day when a regulator asks how a particular decision was made. AI doesn't get me out of any of that. It just changes the surface where the human judgment happens.

The pivot that taught me

My time at Old Well Labs in 2025 was a deliberate, six-month pivot back to IC work for a reason: I needed to know how AI-native engineering practice actually worked at implementation depth, not just at the level of "we should be using more of it." A half-year as staff-level IC at an AI-native fintech will recalibrate any leader's sense of what's hype and what's load-bearing.

The result of that year is that I returned to engineering leadership with a clearer sense of what to ask my teams: where AI is genuinely accelerating us (faster prototyping, better doc generation, fewer "stuck on a regex" hours), where it's actively dangerous (security-critical paths, anything where the model can confidently make up a citation), and where it's a wash (most code review, most CI). Different answers in different domains. That granular taste is the thing my teams pay me for.