2026-09-20 22:01 UTC
@coolboymew@shitposter.world @kumicota@akkoma.kumicota.moe @lain@lain.com @rain@melonbread.xyz right I just think, a smarter runtime LLM translation can build context from what has already been said, and using vision to read the screen. very very solvable.
Replies (1)
-
@coolboymew@shitposter.world 2026-09-20 22:03
@sun@shitposter.world @kumicota@akkoma.kumicota.moe @lain@lain.com @rain@melonbread.xyz True, so it may already be pretty much possible right now if you use the right tool. The issue is self-hosted vs free vs paid (with tokens) and stuff can get expensive