Phase Transitions

AI can do the work.
Choosing what's worth doing is still yours.

More than a year of writing about working with AI, from the cognitive and the practical to the deeply technical. This is the way in.

Subscribe on Substack Read the archive

Start here

New to AI, just curious The strange, tiring feeling of using it all day, named and explained. Why AI Is Tiring You Out ↗︎ You run a team When one person with AI does what a team did, the job of leading shifts. Thinking Like a CEO in the Age of AI ↗︎ You build with it The full operational stack for agents that hold up, with real patterns. Building MCP Servers: Patterns and Lessons ↗︎

By topic

  1. 01

    Getting better results from AI

    Type less, not more. Why shorter, clearer prompts beat careful ones.

    Read on Substack ↗︎
  2. 02

    Giving AI memory and context

    It has a notebook, not a brain. Stop re-explaining; hand it the context on purpose.

    Read on Substack ↗︎
  3. 03

    AI beyond the chat window

    Stop copy-pasting into a box. Give the model eyes and hands on your real work.

    Read on Substack ↗︎
  4. 04

    Making AI trustworthy

    Confidence is not correctness. Build the scaffolding that catches the mistakes.

    Read on Substack ↗︎
  5. 05

    AI as a factory, not a brain

    Not a genius you consult. A line you build around the model, with checks at every stage.

    Read on Substack ↗︎
  6. 06

    Working with your own mind

    Some of the strangeness of AI is about you, not the machine.

    Read on Substack ↗︎
  7. 07

    Leading and building in the age of AI

    Speed is cheap now and direction is not. The operating minds worth borrowing.

    Read on Substack ↗︎
  8. 08

    The bigger picture

    Step back. What is becoming scarce, what is becoming free, and who pays for it.

    Read on Substack ↗︎

Beyond the writing

I build AI systems on the information a business already has, and sit with the people who run the business to decide what those systems should do. Workshops for leadership teams, fixed projects that end with working software, and standing arrangements.

More on the practice