According to gdb, OpenAI achieved super linear unit distances in Erdős’s 1946 problem, with delta≈0.014 per Will Sawin, signaling a math milestone.

According to Greg Brockman on X, GPT-5.4 Pro solved Erdős Problem #1196, with researcher Leeham sharing details and noting that formalization is underway (source: Greg Brockman, original post by Leeham). As reported by the X posts, the result is being verified through formal proof, which is a critical step for mathematical acceptance. According to the posts, if validated, this showcases large language models contributing to open problems in combinatorics, signaling opportunities for AI-assisted theorem proving, automated conjecture generation, and enterprise math tooling in finance, cryptography, and logistics optimization. As noted in the shared thread, community commentary by mathematician Lichtman underscores the problem’s difficulty, highlighting potential business impact for AI vendors offering proof assistants and research copilot products that integrate symbolic libraries and proof checkers.