Essays

Why the integrity of intent through generation is structurally distinct from AI security, safety, compliance, observability, and evaluation, and why treating it as a feature of one of them is a mistake.

The thousand-year genealogy of intent, fidelity, and protocol, and how each word already carried the weight the protocol needs it to carry.

The half of AI verification the compute-and-treaty framing skips: per tool call, per agent action integrity that is buildable now, without waiting for a treaty.