16 Year old used AI to disprove professor

A 16-year-old researcher used a custom GPT-5.5 harness to disprove a case in enumerative combinatorics proposed by Professors Chan and Pak at UCLA and Rutgers in 2024. The work is detailed in arXiv paper 2607.10084, where a supervisor model directed multiple solver agents that reported findings every 90 minutes. The teenager credited Professor Chan for guidance and expressed excitement about future models.

In 2024 Professors Chan and Pak at UCLA and Rutgers proposed a case in enumerative combinatorics. At 16 years old, I used a custom GPT-5.5 harness to disprove this case. arxiv:2607.10084 1/3 - I used a custom harness built around GPT-5.5. A supervisor model directed several solver agents, every 90 minutes, they reported discoveries, failed approaches and remaining gaps. The supervisor then updated their instructions and redirected the search. 2/3Excited to continue working with new and better models on the frontier. Thank you to Professor Chan for the important guidance - arxiv.org/abs/2607.10084 https://arxiv.org/abs/2607.10084 3/3 - Very cool Join the conversation