The Paradigm of Autonomous Research | Francesco Pappone (CEO, Paradigma)

Francesco Pappone (@tensorqt) is building Paradigma - an infrastructure for autonomous research. Their product Flywheel replaces papers with directed acyclic graphs where every experiment is a node with tracked lineage, and AI agents traverse and extend the graph via MCP. In this conversation, we go through Francesco's path from physics to founding Paradigma, how Flywheel works, building from Rome, how it connects to Karpathy's autoresearch and the broader auto-research movement, the hard problems around trust and verification when AI runs experiments, and where autonomous research is headed.TIMESTAMPS00:00:00 - Intro00:01:03 - Dropping Out of a PhD, Saturation in Academia & Building from Europe00:13:15 - AutoResearch, The Idea Behind Flywheel & the ex-OpenAI Co-Founder00:19:18 - Evals for AutoResearch & Research Taste00:25:00 - How Research Labs Use Flywheel00:31:15 - The Thesis Behind AutoResearch & Adoption Across Top Labs00:41:00 - Working with Flywheel & the Combinatorial Explosion Problem01:02:03 - Science as a Public Graph, New Research Loops & Quick Fire01:06:23 - The Toughest AutoResearch Challenge Today & Future Directions01:11:53 - Automating Science: Sakana, DeepMind & the Trust Problem01:21:07 - Business Model, Automating R&D & Model Capabilities01:27:20 - Dropping Out of College to Live on the Frontier