Why does it seem like OpenAI is making scientific breakthroughs, when Anthropic models seem to better?

A user on Reddit observed that OpenAI appears to be making scientific breakthroughs and solving mathematical problems, generating significant news. However, the user's personal experience suggests that Anthropic's Claude models are currently better, despite a lack of headlines about Claude agents achieving similar scientific feats. This raises a question about the perceived difference in public breakthroughs versus actual model performance.