OpenAI Says It Solved Old Problems

OpenAI says its AI models have produced three scientific results that would matter far beyond a product demo: a proof of the local smoothing conjecture in harmonic analysis across every spatial dimension, a disproof of the Exponential Time Hypothesis in computer science, and a mathematically rigorous proof of the existence of quantum chromodynamics in physics.

That is the claim. The evidence available in the published accounts is narrower. Al Jazeera reported the announcement and quoted Alex Iosevich, a University of Rochester mathematics professor who specializes in harmonic analysis, saying OpenAI appeared to have “cracked” the local smoothing conjecture in what he called a historic development. In that field, “historic” is not newsroom confetti. The conjecture is a central proposition, and prior proofs had reached only three dimensions, according to OpenAI.

If that result holds up, it is the cleanest part of the story because the baseline is clear: proved in three dimensions before, now claimed in all spatial dimensions. That is a genuine step change, not the usual AI habit of moving a benchmark by half a decimal point and declaring civilization upgraded.

The other two claims are larger and therefore need even more caution. The Exponential Time Hypothesis is a foundational assumption in complexity theory about how hard certain computational problems are. Disproving it would not be another nice paper for a specialist shelf; it would force a rewrite of some accepted limits in theoretical computer science. Likewise, a rigorous proof tied to quantum chromodynamics touches one of the deepest technical areas in physics. Big claims in old disciplines do not become facts because a well-funded company posts them on the internet.

OpenAI said it worked with outside scientists to turn what it called raw research data into scientific understanding, and that different expert groups chose different release paths. The mathematics and computer science collaborators opted for immediate public release, while the physics group chose to defer publication while writing its own interpretation, according to the company.

That last detail matters more than the marketing gloss. OpenAI is not just claiming results; it is also deciding who sees them first and in what form. The company itself acknowledged that this creates information asymmetry. In markets, people call that price discovery with bad manners. In science, it means the strongest conclusion available for now is the narrow one: OpenAI has announced extraordinary results, and at least one outside mathematician thinks one of them may be real enough to change the field.