Eclipse Corrects Errors in Prior Release on Autoformalization and Mathematical Discovery

Founder Neel Somani reflects on insights from GPT-Erdos experiment exploring AI's impact on mathematical research

Mar. 17, 2026 at 1:26am by Ben Kaplan

Got story updates? Submit your updates here. ›

Eclipse, a blockchain platform founded by Neel Somani, has issued a correction to a prior news release. The corrected release explores how the fusion of artificial intelligence and pure mathematics is accelerating, evolving from computational assistance to enabling creative breakthroughs. Somani recently spearheaded an experiment called "GPT-Erdos" that employed state-of-the-art AI tools to tackle unsolved mathematical conjectures, yielding insights into the informal principles that guide human research and the challenges of defining novelty in an age of autoformalization.

Why it matters

Somani's work highlights how AI systems engaging in "autoformalization" - the process of converting human-readable proofs into machine-verifiable formats - are prompting the scientific community to reconsider long-held notions of novelty, progress, and rigor in mathematical research. As AI-generated solutions often expose ambiguities in how humans define success, it challenges the desire for clean definitions of "newness" and reveals the messy reality of mathematical derivation.

The details

During the GPT-Erdos experiment, Somani observed instances where the AI produced valid solutions that diverged methodologically from established approaches while remaining functionally equivalent. This raises questions about how such results should be categorized - as novel discoveries, rediscoveries, or extensions of prior work. Somani suggests the field may need to adopt a more formalized definition of novelty, potentially measuring the minimum complexity required to express a proof.

  • The GPT-Erdos experiment was recently conducted by Neel Somani and his team at Eclipse.
  • The corrected news release was issued on March 5, 2026.

The players

Neel Somani

A Berkeley-educated computer scientist and the founder of the blockchain platform Eclipse, who spearheaded the GPT-Erdos experiment exploring the impact of AI on mathematical research.

Terence Tao

A leading mathematician whose views on the novelty of AI-generated results may differ from others in the field, highlighting the reliance of the mathematical community on intuition rather than formal logic to assess the value of a proof.

Eclipse

A blockchain platform founded by Neel Somani, which focuses on decentralized technology and sees a direct connection between formal mathematical proofs and software security.

Got photos? Submit your photos here. ›

What they’re saying

“The fusion of artificial intelligence and pure mathematics is accelerating, evolving from computational assistance to enabling creative breakthroughs.”

— Neel Somani, Founder, Eclipse (PR Newswire)

“AI-generated solutions often expose ambiguities in how humans define success. This raises a critical question: How should such results be categorized? Are they novel discoveries, rediscoveries, or extensions of prior work?”

— Neel Somani, Founder, Eclipse (PR Newswire)

“The challenge of defining novelty is not unique to AI; it has long divided even the most accomplished mathematicians.”

— Neel Somani, Founder, Eclipse (PR Newswire)

What’s next

Somani envisions a future where autoformalization incorporates a differentiable surrogate function to measure how close a proof is to being correct, allowing researchers to distinguish between proofs that are fundamentally flawed and those that are nearly complete. This development could transform AI from a binary checker into a true collaborator, capable of navigating the heuristic, iterative process of discovery.

The takeaway

Neel Somani's experiment demonstrates that even if AI progress were to halt today, the practice of mathematics has already been irrevocably changed. The ability to verify proofs via machine and rapidly assimilate existing approaches allows researchers to focus on high-level conceptualization rather than rote memorization, redefining the very nature of the questions they seek to answer.