Technical note
Typed Authority Boundaries
How technical artifacts can render, translate, validate, and report on authoritative structures without silently acquiring the power to redefine them.
Semantis Research
Research for systems that must earn authority.
Semantis Research investigates the mathematical, computational, and methodological foundations required for intelligent systems to remain safe, accountable, and recoverable as capability becomes consequence.
Semantis Research / 2026
The program
Safety is not a property of the model alone.
Modern AI systems increasingly reason, invoke tools, delegate, and interact with external systems. Semantis studies the structures required to determine when information, components, permissions, and actions may legitimately enter trusted operation.
The broader program spans safety and assurance, verification, authority, admissibility, recovery, and the mathematical foundations required to reason about them rigorously.
Public research
Selected work from a larger research program.
Public research is intentionally selective. The objective is to make important claims examinable without publishing the complete architecture required to reproduce the underlying systems.
Methodology
The Burden of Inquiry
Why inquiry carries obligation, why conclusions must preserve the reasons by which they survived, and why inherited authority requires recoverable justification.
Research statement
Earned Authority in AI Systems
Why generation, evidence, evaluation, certification, permission, and authority should remain distinct as AI systems become more capable of consequential action.
Research posture
Engagement
Selected engagement is welcome.
Semantis Research is exploring selected funding, evaluation, research, and strategic relationships. Initial discussions remain non-confidential; deeper technical disclosure is reserved for appropriate diligence.