According to CoinMeta, OpenAI refers to a secret AI model that wrote 722 mathematical papers, almost all of which originated from a single hint. This phenomenon was cited by Ethereum researcher Justin Drake, who warned that AI might be able to crack Bitcoin wallets. This unpublished model was assigned about 4000 open problems, and OpenAI only retained the results that it deemed important. Of the 722 papers, only 162 have main results that have been computer-verified, and OpenAI acknowledges that "some unformalized results may have issues," with the specific details coming from Decrypt.