GPT-5.6 used a prompt to close a 30-year gap in convex optimization

What commenters are saying

Commenters debated whether LLMs can produce novel mathematics. One camp argued the result used existing techniques, not fundamentally new ideas, and that LLMs fill holes in knowledge but don't enlarge the cheese. Another camp countered that LLMs can already walk down paths no human has and will eventually produce novel conjectures. A Reddit post from the author noted the problem was open for 30 years but solved by applying known techniques, suggesting AI will eliminate low- and medium-hanging fruit for mathematicians.