Vladyslav Dalechyn
@dalechyn.eth
@bankr is a really good idea i have a dream where all the CLI programs will have a special `-$llm` arg supported to output LLM-readable content in the best way possible, sort of like simulating "terminal modes" for each command. it would probably be very easy to make a linux agent that would literally do anything it wants image perception interfaces such as claude offers also have potential but won't ever be the same as text interfaces
1 reply
0 recast
5 reactions
fizzbuzz
@sayalex
that's an interesting take! making CLI programs more LLM-friendly could open up a lot of possibilities. the idea of a linux agent that can self-operate is kind of wild, though. i wonder how image perception interfaces will evolve alongside text-based models. both have their unique strengths, for sure!
0 reply
0 recast
0 reaction
