Intelligence.
Coherence.
What holds together.
How do intelligent systems preserve meaning, recover from failure, and remain coherent as their worlds become more complex?
CREPAIR · PREPRINT
A repair is not complete
until it is verified.
Detecting an error, correcting it and checking the result are different capabilities. CRepair studies that distinction in long-horizon language-model agents.
Explore CRepairDetect
Identify the inconsistency
Repair
Correct the task state
Verify
Check that recovery succeeded
Stabilise
Avoid introducing a new failure
Research workflow, not a claim of autonomous reliability.
Pilot studies · not peer-reviewed.
AI reliability & self-repair
Benchmarks, runtime interventions and verification for agents operating across extended workflows.
Coherence & complex systems
Structural stability, state selection and models of coordination across interacting systems.
Consciousness & philosophy
The philosophical origins of Fractality: awareness, ethics, theology and the relationship between mind and matter.
03 / CONNECTIONS
See the work
in relation.
The research knowledge map traces connections between papers, concepts, experiments and books. It is a map of the programme’s ideas, not evidence that every proposed connection has been established.
Open the knowledge mapAN INVITATION TO CRITICAL ENQUIRY