Hacker News new | ask | show | jobs
by spongebobstoes 37 days ago
newer models tend to use fewer thinking tokens to solve the same problems, and is a strong counterexample to your entire comment