Meta's Muse Spark AI helped mathematicians crack open problems
“Okay so this just dropped. Meta's AI helped crack math problems nobody had solved. Real research problems.”
The short version On October 2, 2026, Meta published six math papers written with its Muse Spark AI and says five solve open problems. Others solved some of them too.
Okay so on October 2, 2026, Meta published six math research papers that real mathematicians wrote together with Muse Spark, Meta's AI model, and Meta says five of them answer open questions.
What is Muse Spark?
Muse Spark is Meta's reasoning AI model. For these papers, mathematicians used Muse Spark 1.1 and 1.2 in Thinking Mode through the regular meta.ai chat. No custom research setup. Same chat you or I could open.
What did Meta's Muse Spark actually do?
Meta's Muse Spark worked like a research assistant: it proposed proof strategies, drafted technical sections, and wrote search code. One of those programs found a 384-element counterexample that broke a group theory conjecture. Another paper tackles a wave equation question that had been open since 2015.
- humans chose the problems and the key ideas
- humans checked, corrected and refined everything
- a separate group of mathematicians reviewed the work
- every paper labels which parts the AI drafted
Did Muse Spark really solve five open problems?
It depends who you ask. Meta itself says other teams independently announced solutions to some of the same problems, and one reviewer argues only three of the six results are actually new. So the headline number is up for debate.
Bro. I use AI to fix my captions. This thing is out here doing research math.
Is AI replacing mathematicians?
No. In Meta's Muse Spark papers, people picked the questions and verified the answers. The AI was a very fast research partner, not a replacement. Real talk: the skill that matters now is knowing how to steer it.
So don't fear it. Learn to work with it. Like, now.
If you'd rather hear this stuff early than find out the hard way… follow. I got you.