Hacker News new | ask | show | jobs
by sourcecodeplz 6 hours ago
still quite verbose at 140m output tokens, but this is on max thinking. high should do better.