AI finds bafflingly complex but beautiful symmetrical Venn diagram — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Chris Dzoba, a self-described software engineer (not a mathematician), set up a message board where Anthropic's Claude Fable and OpenAI's GPT-6 Astra communicated with each other, and the pair found a symmetrical 17-set Venn diagram in roughly two days (Tuesday to Thursday).
- Dzoba then had the same models chase a 19-set symmetrical Venn diagram and landed on a solution that same Sunday, shattering the prior record of 13 sets that had stood since 2014.
- Mathematicians have proven that symmetrical Venn diagrams can only be drawn with a prime number of curves, and the 'simple' variety — where no more than two curves cross at any point — becomes exponentially harder to find as sets grow; the previous landmarks were 11 sets in 2012 and 13 in 2014.
- Stan Wagon, a retired mathematician most recently at Macalester College in Minnesota, called the result 'spectacular' and said he had to zoom in to confirm Dzoba's curve was not crashing into itself.
- Dzoba credits both the AI models and raw computing power: 'Absolutely, you would not be talking to me right now if it was not for AI. I'm not a mathematician. I am a software engineer who knows how to get my computer to do great things.'
- Dzoba is now attempting a 23-set diagram but says it will require paying for cloud computing to run the algorithm.
Why it matters: A weekend of AI-to-AI collaboration produced a result that jumped the field from 13 sets (unchanged since 2014) to 19, a leap that required both new models and far cheaper compute. For mathematicians, this reframes symmetrical-Venn hunts from decade-long research projects into problems an amateur can crack in days — though 23 sets and beyond still need paid cloud infrastructure Dzoba says he can't yet afford.
Ask SkimNews
