This website documents a number of conversations with GPT-4, a model released by OpenAI.

Note that sometimes the “user” in the conversation is me, but not always. To scale up the number of conversational topics I could explore, sometimes “user” is either a character who is also simulated by GPT-4 or mimicks a template.

Please also note that the conversations with the model relating to scientific hypothesis generation have not been fact-checked by domain experts. Nor have the simulation predictions been verified. Therefore, you should not assume that they are free from error. If you are a domain expert and you have comments/feedback on one of the conversations, I’d be happy to hear it. Nevertheless, I hope the samples will prove of interest.

Contact Samuel ( for questions. Credit to Vlad for producing the imagery.