Skip to content

Comment on Quantitative AI progress needs accurate and transparent evaluationparent

Comments

LLMs are better at LaTeX than humans. ChatGPT often writes LaTeX responses.

Yeah, it's honestly one of the things they're best at!

I've been working on implementing some E&M simulations with Claude Code and it's so-so on the C++ and TERRIBLE at the actual math (multiplying a couple 6x6 matrix differential operators is beyond it).

But I can dash off some notes and tell Claude to TeXify and the output is great.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.