AI models struggled with high-school math in 2023, did well on top high-school competitions in 2024, solved minor open problems in 2025, more significant ones in early 2026, and in late 2026 are beginning to solve top open problems in mathematics.
Unresolvable The public record could not settle it.
Factual: Statements about the present or the past, checked against the record: did they hold up?
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
The claim lists four dated milestones for AI in mathematics: (1) in 2024, did well on top high-school math competitions (for example IMO-level results); (2) in 2025, solved minor open problems; (3) in early 2026, solved more significant open problems; (4) by late 2026 (as of 2026-09-23), beginning to solve the top few open problems in all of mathematics (for example Millennium Prize Problems, the Riemann hypothesis, or problems of similar stature). The proposition also says AI struggled with high-school math in 2023. TRUE: all milestones are documented in their stated periods, including credible, expert-verified AI progress on at least one of the very top open problems by 2026-09-23. MOSTLY TRUE: all milestones hold, with a minor slip in timing or degree, but some credible progress on a top-tier problem still exists. MIXED: the earlier milestones (2024 competitions, 2025 minor problems, early 2026 significant problems) hold, but there is no credible evidence that AI was beginning to solve the top few open problems by late 2026. MOSTLY FALSE: only the competition milestone and perhaps minor open problems are verified, and the later milestones are not. FALSE: most of the milestones did not occur in the stated periods.
Set before the research began.
Evidence (3)
-
E1 Anthropic CEO says AI could cure most major diseases within next decade Fox Business · 23 September 2026
“In 2024, they started to do well on math competitions for the best high-schoolers in the country, in 2025, they started to solve minor open problems, in early 2026 more significant open problems, and in late 2026, they are beginning to solve the top few open problems in all of mathematics.”
-
E2 25 winners of math’s ‘Nobel Prize’ decry the AI invasion of their discipline Scientific American · 14 September 2026
“last Tuesday OpenAI claimed a solution to the Navier-Stokes equations, one of the seven “Millennium Prize Problems,” considered the subject’s hardest open questions.”
-
E3 25 winners of math’s ‘Nobel Prize’ decry the AI invasion of their discipline Scientific American · 14 September 2026
“This human culture is being endangered by the “mass production” of AI-generated proofs, it adds.”
Reasoning
E1 only restates the claim. E2 says OpenAI claimed to have solved the Navier-Stokes Millennium Prize Problem, but it reports this as a claim and gives no expert verification. E3 mentions the "mass production" of AI-generated proofs, but it does not tie that to any of the dated milestones. None of the evidence covers the 2023, 2024, 2025 or early-2026 milestones, and none shows verified progress on a top open problem, so the criteria cannot be applied.
Cite this claim
Dario Amodei, “AI models struggled with high-school math in 2023, did well on top high-school competitions in 2024, solved minor open problems in 2025, more significant ones in early 2026, and in late 2026 are beginning to solve top open problems in mathematics.” (said 23 September 2026). Verdict: unresolvable. Reliability API, claim c_ff7cbe0b8af7, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_ff7cbe0b8af7
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Dario Amodei
-
Mixed
“It’s easy to dismiss this as a one-off or curiosity, but we’ve repeatedly seen a pattern where AI performance in new intellectual domains goes from weak to superhuman in a matter of a few years.”
-
Mixed
“we were measuring these trends back when models cost $1,000 or $10,000 to train”
Amodei's group measured AI scaling trends when models cost $1,000 or $10,000 to train.
-
Mixed
“a lot of the other reasoning models in the market are trained primarily on math and competition coding,”
-
Mostly true
“I do think AI coding is moving the fastest of all the other areas.”
Amodei said he thinks AI coding is moving the fastest of all the areas of AI.
-
Mostly true
“Because when it knows that it doesn't get any gain out of thinking further, it doesn't think for longer.”
Verdicts are made automatically; how verdicts are made.