Writing

Notes on private AI, from the people building it.

• July 17, 2026

  Browser Use and an Agentic Marketplace

  We rearchitected our Browser Use agent: bring your own OpenAI-compatible endpoint, no master plan, and an extensible repository of skills. Where it leads is a marketplace of digital humans bidding on tasks.

• July 13, 2026

  The Reverse Information Paradox

  Every time your firm uses cloud AI, it teaches the provider how your firm works. In the age of intelligence, protecting your core IP means owning the machinery your firm learns with.

• July 13, 2026

  Your AI policy is working. That's the problem.

  Your AI policy is working. That's the problem.

• July 12, 2026

  A Big Brain and Many Small Hands: How We Cut Browser Agent Costs by 50x with a Local-First Architecture

  Pair one large cloud model that plans, delegates, and verifies with a set of small local models that do the grunt work. The large model thinks a few times per task. The small models act dozens of times per task, at zero marginal cost.

• July 11, 2026

  How much is AI really costing your business?

  Cloud AI could be costing your business more than just the subscription fees

• July 11, 2026

  The case for running your own local AI stack

  Why depending on a few frontier AI providers is a fatal weakness—and how small, sovereign models you own and run change the game.