Exploring alternative architectures, inference systems and model behaviour.

Experimental work developed through controlled testing, comparative benchmarks and falsifiable claims.

Evidence over narrative. Claims remain provisional. Negative results remain evidence.

Current Work