The Thread
At the University of Milan, I studied epistemology — the branch of philosophy that asks not what we know, but how we can know anything at all. From Popper's falsificationism to Russell's logical atomism, from Kuhn's paradigm shifts to Lakatos's research programmes: the rigorous analysis of how humans construct, validate, and revise scientific knowledge.
At Harvard, working alongside John Nash on game-theoretic equilibria and Gary Chamberlain on econometric inference, I saw these philosophical questions become computational ones. How do we reason under uncertainty? How do we update beliefs with new evidence? How do we distinguish signal from noise in complex systems?
This is why my AI systems are different. They're built on an epistemological foundation: every claim must be traceable to evidence, every inference must be auditable, and the system must know — and communicate — the limits of its own knowledge.
Most AI today hallucinates confidently — generating plausible-sounding answers with no grounding in verified knowledge. This isn't a bug to be fixed with better training. It's a fundamental architectural failure: systems built to predict the next token, not to reason about truth and justification.
PRISM takes a different approach. It creates Professional AI Twins grounded in domain-specific knowledge bases, with explicit reasoning chains, source citations, and epistemic boundaries. Not "AI that sounds smart." AI that knows what it knows — and tells you when it doesn't.