
0
0
Egypt
3+ years of experience
LLMS Do Make Mistakes. This is the reality we cannot deny. We can overcome this issue with our understanding, prompts and Agentic.md Files. Asking LLMS to Count letters is Stupid, asking them to create a python app which does that is perfect. The same model answering the same prompt correctly before a new release and incorrectly after proves silent degradation by the provider, which introduced a regression into the model. I’m fully prepared to admit when I’m wrong, hopefully that won’t happen too often…