Dev argues prompting alone can't get AI to solve natural science problems

felpix_ · x · 2026-09-22

Responding to OpenAI's claim that its model solved "100+ long-standing open problems" in math within two weeks of the NS announcement, developer felpix argues on X that no average person—or even a 90th-percentile user—could prompt the model to solve natural science problems. Beyond cost and compute, you'd need a team of PhD experts who know the state of the literature and which ideas have promise—which, he notes, is exactly the team OpenAI assembled.

Related event: Developer Questions Whether Ordinary Users Can Solve Science Problems via Prompts(3 posts)→

Original post →

More from Models

Models channel →