I use this space to make the thinking behind my work visible: what I’m learning, what I’m testing, why I choose one approach over another, and what evidence would make me change my mind. Some pieces are finished essays. Others are active research or experiments.
AI can generate a lot of the implementation. I’m more interested in the decisions around it: the constraint, the tradeoff, the failure mode, the evidence, and what would make me choose differently.
Why I use an intentionally boring format for event history, auditability, replay, portability, and separating historical evidence from mutable application state.
Read the essay →Why I prefer HTML-first, server-owned workflows when they fit—and why simpler architecture still has to survive real security and reliability requirements.
Read the essay →I’m interested in the contract that already exists before I invent new behavior: state ownership, retries, compatibility, failure semantics, and trust boundaries.
Active researchRight now I’m spending time on attention and judgment in AI-assisted work; how humans supervise multiple agents without micromanaging them; memory versus attention; backpressure and escalation; contract-first system design; and how evidence, replay, and observability make automated systems easier to understand and trust.
Not every idea here is a finished position. I want this section to show the work while it is becoming clearer, not pretend every question already has an answer.