Totally agree. We’ve seen similar weirdness when trying to build deterministic behaviors around LLMs. It’s fun at first…. Until you’re debugging something that

Totally agree. We’ve seen similar weirdness when trying to build deterministic behaviors around LLMs. It’s fun at first…. Until you’re debugging something that just needed a if/else. We’re now mixing prompts with conventional logic exactly for that reason, LLMs are powerful, but not magical.