Six brand new mathematical discoveries were generated by an autonomous multi-agent system with zero human intervention.
April 23, 2026
Original Paper
Bolzano: Case Studies in LLM-Assisted Mathematical Research
arXiv · 2604.16989
The Takeaway
AI is moving beyond the role of a basic calculator to become a creator of new knowledge. This multi-agent system produced publishable results in theoretical computer science. Three of these findings were completely novel and not just refinements of existing work. It proves that the reasoning gap in AI is closing faster than experts predicted. Mathematical research is no longer the exclusive domain of human intuition and grit. We can now deploy fleets of agents to hunt for proofs in areas humans have not yet explored.
From the abstract
We report new results on six problems in mathematics and theoretical computer science, produced with the assistance of Bolzano, an open-source multi-agent LLM system. Bolzano orchestrates rounds of interaction between parallel prover agents and a verifier agent while maintaining a persistent knowledge base that is carried across rounds. Classified using the significance-autonomy taxonomy of Feng et al., four of the six results reach the level of publishable research, and three of the six were pr