I would not have trusted LLMs with C code a year ago but today's models appear to be very good at reasoning through memory management and other tricky aspects That said I'm not enough of a C expert myself to credible evaluate what they're doing!

→ View original post on X — @simonw