A very basic litmus test for LLMs "ok give me a python program that reads my c: and put names and folders in a sorted list from biggest to small"

Then ask your cloud FOTM api to verify the code it spit.
I thought it was an easy question, but my local ones just died on it, with wrong executions, double-reading the sizes of files, putting recursive functions inside recursive functions.

I think I got my magic test.