Writing / Field notes and essays

Thinking in public, close to the work.

On engineering leadership, practical data and AI, and lessons from coordinating teams of AI agents.

Subscribe via RSS

The archive

All writing

  1. Data and AI, Engineering leadership

    A prototype is not a product, even when AI built it

    AI lets anyone build a working tool. The gap between a tool that works for its author and one others can rely on is still engineering, and the scarce skill is choosing which to invest in.

  2. Agentic engineering, AI agents, Measurement

    An agent budget is a scope decision

    Telling an agent to stop at a token count did not hold for us. Sizing the task so an overrun is cheap, and measuring spend outside the agent, did.

  3. Engineering leadership

    When being the problem solver becomes the problem

    Jumping in to fix the hard thing feels like leadership, but it often makes a team more dependent on you. A short way to decide when to step in and when not to.