@w32ksys@undercoversign@neptune_insider Local models are extremely underpowered compared to big cloud models, you'd just be disappointed; use a provider that hosts a big model like Kimi K3.1 Pro or something
@jrysana@VictorTaelin I think thats the best way too; I think Taelin is extremely brilliant, and it would be much better for him to focus on what is his calling and what he knows best (research), and let someone else, whos a bit more "salesy", do the raising, so Bend gets the best of both worlds.
@caffeinum@thdxr The plan for yapfast is to have like "adapters", which could support stuff like instagram, whatsapp, and at some point maybe even Slack; so it's not totally off the table:p
@MaleManlpulator@thdxr They did; buuut it will be closed source, takes 300MB of RAM still, and won't support plugins (which yapfast is going to have very soon) :p
@mdmaenthusiast@thdxr Last time I tested this it wasn't the case, since the backend (skia) didn't seem to work that way; I can retest though to be sure.
@mdmaenthusiast@thdxr Oh yeah totally; for high resolution screens it doesn't make much sense, but for GPUless or old GPUs on smaller monitors it saves around 80MB RAM (not VRAM) that you have to dedicate towards GPU rendering.
I am! I'm counting overall system memory usage:p
@thdxr Trading off potentially slower renders and more CPU time for less memory usage, if the user chooses so; with https://t.co/yTq7pr9JST
Right now I use Slint (which supports both CPU and GPU backend), and it's good, but the compile times are a nightmare :D