2026-04-27 22:34 UTC
That's pretty cool and makes a lot of sense.
@screwlisp
Replies (1)
-
@screwlisp@gamerplus.org 2026-04-27 22:43
@dogfox I remember that prior to this, some people were saying that while right now LLMs are apparently steered by "prompt engineering", later it was going to turn out that they are better steered through line noise sequences of characters. Which I guess this discovery confirmed. It does not seem to have been picked up by main stream discourse yet. I guess it sucks for the vendors that the Natural Language Prompting thing was a sham.