There is a sentence I never expected to write on this blog: an AI company is claiming new results in pure mathematics and theoretical physics, not as a demo, but as contributions that working scientists can verify. OpenAI says its models have produced results touching the Kakeya conjecture in mathematics, the Exponential Time Hypothesis in theoretical computer science, and problems in quantum chromodynamics, the physics of the strong nuclear force.
I need to handle this story the way this blog always handles big claims: with respect for what it would mean, and with both feet on the ground. These are not product benchmarks. The Kakeya problem has occupied serious mathematicians for decades. The Exponential Time Hypothesis sits underneath huge parts of computer science. If an AI system genuinely moved either one, it would be a landmark in the history of research, not just of AI.
And that is exactly why the next step matters more than the announcement. Results like these live or die in verification: independent experts reading the arguments line by line, hunting for the hidden assumption, the quiet gap. OpenAI is presenting at least some of this work in forms researchers can examine, which is the correct move. Extraordinary claims are welcome in science. They just do not get to skip the queue.
Here is the part I find genuinely historic, whichever way the checking lands. We have crossed from AI that summarises what humans know into AI that proposes things humans do not yet know, in fields where being wrong is not an opinion, it is a proof error. A chatbot can flatter you. A proof cannot. That makes mathematics the most honest arena AI has ever stepped into, and I love that the arena does not care about anyone's valuation.
For readers of this blog, the practical meaning is simple. The ceiling on what these tools are for just moved again. Yesterday they were for drafts, designs and code. Now serious people are using them at the edge of human knowledge, and the tools on your desk are cousins of the ones doing it.
I will follow the verification on this blog as it comes in, and if the results fall, I will write that too. But if even one of these claims stands, remember this week. It is the week AI stopped only answering our questions and started handing us its own homework to mark.
Haroon Ch
Ai Game Changer Pro
