Prompt issue

[nmlemus]

“Smarter” models such as Claude Sonnet, Gemini 3, or GPT-5 correctly understand the prompts and perform the intended actions, while other models fail to do so. (example: the decision to use the /data folder, the use of latex, etc)