AI just solved 10 open math problems no human had cracked should we be excited or worried?

Saw this making rounds today: an internal OpenAI model reportedly solved 10 previously unsolved problems in math and theoretical CS, and actually published verifiable proofs for them. Not a benchmark flex, actual open research territory.

As someone building in this space, it's wild to watch the line between "good at test-taking" and "doing real research" blur this fast. A few things I keep thinking about:

• If models can generate verifiable novel research, what does that mean for R&D teams, especially small startups without a research bench?
• Does this change what "moat" even means for AI-native products?
• How many of you are already using frontier models for actual research work vs. just code/content?

Curious what this community thinks: hype cycle or genuine inflection point?