Eclipse Research Explores AI's Role in Mathematical Discovery

Blockchain platform founded by Neel Somani examines how advanced AI can assist in solving complex math problems through autoformalization.

Published on Mar. 5, 2026

Got story updates? Submit your updates here. ›

Eclipse, a blockchain infrastructure platform founded by computer scientist Neel Somani, has released new insights into the growing role of artificial intelligence in mathematical discovery. The research initiative examines how advanced AI systems can assist mathematicians in solving complex problems through a process known as autoformalization, the conversion of human-readable proofs into machine-verifiable mathematical formats.

Why it matters

The research highlights how AI-generated proofs can expose ambiguities in how mathematicians define success, novelty, and originality. As AI systems increasingly generate mathematical proofs, the traditional boundaries between rediscovery, incremental improvement, and genuine innovation are becoming less clear, requiring new ways to measure the significance of AI-assisted research.

The details

The project, known as GPT-Erdos, involved applying advanced AI models to open mathematical problems originally posed by mathematician Paul Erdős. The experiment produced partial solutions, rediscoveries of previous results, and new perspectives on existing techniques. Eclipse researchers observed that AI-generated proofs often satisfy formal requirements while diverging from human expectations, highlighting the challenge of underspecification in mathematical problems.

  • The GPT-Erdos experiment was conducted in 2026.

The players

Neel Somani

The founder of Eclipse, a blockchain infrastructure platform, and the driving force behind the research initiative exploring the role of AI in mathematical discovery.

Paul Erdős

A prolific mathematician whose open problems were used in the GPT-Erdos experiment to test the capabilities of advanced AI models.

Got photos? Submit your photos here. ›

What’s next

Eclipse is exploring approaches that could introduce new metrics for measuring the "closeness" of a mathematical proof to completion, allowing AI systems to identify promising proof paths even when a fully formalized result has not yet been achieved.

The takeaway

The research from Eclipse highlights how the intersection of artificial intelligence and formal mathematics may fundamentally reshape the practice of mathematical discovery, with AI systems acting as collaborators capable of exploring vast conceptual spaces and rapidly verifying complex reasoning.