Undomesticated AI

Long-form / arguments / architecture

Essays.

Arguments that need enough room to expose the claim, the pressure against it, and the evidence boundary.

Published

Arguments admitted to the public record.

Editorial rule: an essay may contain speculative candidates. It must not disguise them as implemented, validated, externally proven, or production-ready facts.

Planned lines

What belongs here.

  • Capability without inherited authority.
    Why stronger machine competence does not itself establish lawful execution rights.
  • Persistent AI as infrastructure.
    What changes when AI is not a tool opened for a task but a continuous participant in work.
  • The evidence problem in agentic systems.
    Why the most important output may be the receipt around an action rather than the action alone.
  • Discovery before domestication.
    How premature category assignment and forced practicality can erase genuinely new candidates before they reach falsification.