Observable AI cognition
Cognition is studied through measurable runtime state rather than hidden model weights alone.
Cognitive Runtime Lab
The unit behind the project.
Cognitive Runtime Lab is an independent research unit studying externally observable AI cognition. Its work focuses on runtime attractors, measurable cognitive dynamics, safety arbitration, operational concept formation, and reproducible experiment pipelines.
ADAM as first testbed
ADAM is the lab's first research system for attractor dynamics and concept formation.
Evidence before claims
Claims are expected to survive replay, ablation, cross-seed checks, and artifact audit.
People
Small by design, evidence-first by default.
Lab Principle
Build systems that can be inspected before they are believed.
Cognitive Runtime Lab is organized around one public standard: cognitive claims about AI systems should be tied to observable state, repeatable evidence, and careful limits.
Identity
Independent research unit
The lab develops the Cognitive Runtime Project and ADAM research system.
Contact
elroy@cognitive-runtime.org
Research correspondence, collaboration inquiries, and artifact requests can be directed here.