Make machine action legible before output becomes irreversible consequence.
PUBLIC PROGRAM / SELECTED RELEASES
The Tell
AI hidden states · verifiable agency
“What the model knows before it says it.”
Auditable agent architectures, typed commitments, reversible cognition, hidden-state probes, and cryptographic continuity.
ActiveTyped epistemic posture, claim settlement, state continuity, adversarial traces, verifier independence, and evidence-linked memory.
01Seven-Verb framework
02RBSA
03Claim lifecycle
04Hidden-state measurement
Public material is limited to the surface that can be stated precisely without exposing protected implementation, sensitive relationships, unfinished claims, or invention strategy.
PROGRAM POSTURE
MEASUREMENT-FIRST AI
ACTIVE BUILDThe program turns agent behavior into typed, auditable commitments with cryptographic state continuity, so that what a system did can be shown rather than asserted. The governing position is plain: a system you cannot measure is one you cannot trust or certify. Method and claim language are held at controlled aperture while the invention estate is on file.
ASSOCIATED RESEARCH
3 LINKED RESEARCH FILES.
Program relationships are canonical register bindings. They do not imply publication, outside validation, or unrestricted access.
GY-R-001
Typed Epistemic Commitment for Verifiable AI Agent Behavior: A Seven-Verb Framework with Cryptographic State ContinuityArchitecture paper · Review candidate
Controlled access↗GY-R-020RBSA: A Reversible Belief-State Architecture for Auditable Agency, Counterfactual Retention, and Selective CommitmentArchitecture paper · Review candidate
Controlled access↗GY-R-025Order Effects and Non-Classical Contextuality in Language ModelsMeasurement framework · Open publication candidate
Public briefing↗