Principles

Principle 1 — Interpretation is a governed surface.
Models interpret. Governance defines what counts as acceptable interpretation in a given context.

Principle 2 — Claims must be separable from inferences.
Separate what is asserted, what is derived, and what is unknown.

Principle 3 — Constraints precede optimization.
Safety, legitimacy, and scope constraints come before performance targets.

Principle 4 — Auditability is not optional.
If it cannot be audited, it cannot be trusted at scale.

Doctrinal posture

Doctrine can be public, execution can be private.
Machine-first surfaces should reduce ambiguity, not add noise.
Scope boundaries prevent interpretive drift.