Writing · Field notes

Notes on building with AI twins.

Field notes on engineering, governance, and what changes when a twin does the work.

  1. Strategy

    The Data Sovereignty Flywheel

    Owned operational data gives agents richer context, and that context compounds into better work and less vendor dependence. Jun 23, 2026 · 6 min read
  2. Strategy

    The Shifting Calculus of Build vs. Buy

    Coding agents are reducing both the cost of building software and the long maintenance tail that once made buying the safe default. Jun 23, 2026 · 5 min read
  3. Open Source

    Contribution Without Pull Requests

    Duaility open source projects will not accept pull requests. We still want contributors; we want their judgment where it matters most. Jun 19, 2026 · 6 min read
  4. Engineering

    An AI Twin Is Not Just an Agent

    An agent performs tasks. An AI twin adds durable constraints, memory, checks, and accountability around that work. Jun 19, 2026 · 4 min read
  5. Engineering

    Introducing Governance Kit

    Governance Kit is the open-source layer that turns repeated agent corrections into repo-native rules, checks, and receipts. Jun 19, 2026 · 4 min read
  6. Engineering

    The Verification Loop

    Testing used to give me confidence in my own code. With coding agents, that assurance had to become a standing verification loop. Jun 19, 2026 · 3 min read
  7. Engineering

    The Work Is the Same. The Working Style Is Not.

    Humans and agents can work on the same feature, but continuity, parallelism, patience, and interfaces make them effective in different ways. Jun 22, 2026 · 7 min read
  8. Engineering

    The Inversion of Software Engineering

    Coding agents make production cheap. The harder shift is moving from adding code to living in default refactoring mode. Jun 12, 2026 · 4 min read