A lab announced that its AI models achieved gold-medal-level performance across five competitions spanning mathematics, physics, and chemistry. It frames the next question as whether AI can contribute to genuinely open problems with no existing solution path, citing work over the past several months.
OpenAI published ten results on long-open math problems, each with reasoning walkthroughs, claiming its models produced advances where no progress had been made for at least a decade. The posts focus on the validation process for AI-assisted discoveries, questioning what proof standard should be required before such results are accepted as mathematical knowledge.
OpenAI published ten results on long-open math problems, each with reasoning walkthroughs, claiming its models produced advances where no progress had been made for at least a decade. The posts focus on the validation process for AI-assisted discoveries, questioning what proof standard should be required before such results are accepted as mathematical knowledge.
OpenAI revealed it produced solutions to 10 long-standing mathematics problems, some unsolved for decades. Oxford Fields Medalist James Maynard told The Verge he has been 'soul searching' about the field's future as it adapts to AI.
OpenAI revealed it produced solutions to 10 long-standing mathematics problems, some unsolved for decades. Oxford Fields Medalist James Maynard told The Verge he has been 'soul searching' about the field's future as it adapts to AI.
GPT 5.6 Pro solved all six problems from the 2026 International Mathematical Olympiad on its first attempt, without human help or steering. Former IMO medallists assessed the work and published a report, marking a notable milestone in AI mathematical reasoning.
GPT 5.6 Pro solved all six problems from the 2026 International Mathematical Olympiad on its first attempt, without human help or steering. Former IMO medallists assessed the work and published a report, marking a notable milestone in AI mathematical reasoning.
A Reddit user posed a question about the capability gap between AI solving preexisting open problems and AI independently venturing forward in mathematics. The thread, posted on August 5, 2026, seeks input from high-level mathematicians on this distinction.
A Reddit user posed a question about the capability gap between AI solving preexisting open problems and AI independently venturing forward in mathematics. The thread, posted on August 5, 2026, seeks input from high-level mathematicians on this distinction.
Anthropic announced that an unreleased model made progress on the Riemann hypothesis, a 150-year-old unsolved math problem with a $1 million bounty. The finding could reopen questions about AI's ability to discover new mathematical ideas.
Anthropic announced that an unreleased model made progress on the Riemann hypothesis, a 150-year-old unsolved math problem with a $1 million bounty. The finding could reopen questions about AI's ability to discover new mathematical ideas.