[pipeline_peak]

> AI, as it stands, screws up the basics, let alone something if this scope

Do you have examples?

[Novosell]

Ask the LLM for examples of LLMs fucking up on simple tasks. Either it succeeds, proving the point, or fails, also proving the point.

I had both GPT-4o and llama3.1, through duck.ai, make up kscreen-doctor commands the other day. Commands that were easily formatted by simply looking at the output of kscreen-doctor --help.

[shakna]

Pretty much everything shown here [0]

[0] https://arxiv.org/pdf/2410.05229