Newer Claude Models Are Worse at Using Your Tools: An Engineering Trap of RL Training Backfire
Armin discovered that Opus 4.8 and Sonnet 5 are actually more error-prone than older models when calling Pi's …
Armin discovered that Opus 4.8 and Sonnet 5 are actually more error-prone than older models when calling Pi's …
Claude can't do math or query databases on its own. But you can teach it to 'make phone calls' — give it the …