[zbendefy]

Good point!

There is also the case for Markov chains being theoretically able to do these if tuned well. Or even SAT problem.

[CGMthrowaway]

"LLM is just fancy autocomplete"

[galaxyLogic]

LLM is an Oracle