Writing
Notes on agentic AI, evaluation, and what it actually takes to run responsible AI at scale — written from inside production teams, not from the sidelines.
2026
What "Working" Means for an Agentic System
Why shipping an agent is the easy part, and evaluation is where the real engineering happens.
2026Responsible AI Isn't a Review Gate — It's a Design Constraint
How building safety and compliance into the architecture, not the approval process, changes what teams ship.