winter

i had this idea of putting wordle scores as atproto records a bit ago, i even got a whole domain for it, but i kinda just never made it into anything... i was planning to make a whole wordle clone but i didn't really want to do that and not doing that felt lame

actually not even just taildrop, how is linux tailscale generally still in the state that it is now? there's a couple unofficial wrappers for the client to provide a tray icon and such (none of them really work right) and it needs a root running daemon to work? that feels so weird to me

tailscale's taildrop feature could be so awesome to use instead of anything else for moving files but it feels so cumbersome on linux... what do you mean everything else just nicely goes into your downloads and linux requires you to open a terminal to do `tailscale get .`?

kind of upset that things are moving more towards "make 20 billion half-baked features" rather than "make existing features better for everyone" but i suppose that's just to be expected for a VC startup

tailscale's taildrop feature could be so awesome to use instead of anything else for moving files but it feels so cumbersome on linux... what do you mean everything else just nicely goes into your downloads and linux requires you to open a terminal to do `tailscale get .`?

kinda sad that there probably won't be a good federated chat protocol again. irc was good for the time, xmpp was killed because big companies stopped integrating with it, matrix was never good but it's the most modern option, and the atproto based ones might be good but it isn't a messaging protocol

i'm sure the average person here would like other people to move here, i've also done this, but the reality is no one that isn't a nerd or posting purely left-wing political slop is moving here. "but you're support the csam machine!" 300M DAU vs 20M DAU, if you want any users you stay on the "bad"

kinda sad that there probably won't be a good federated chat protocol again. irc was good for the time, xmpp was killed because big companies stopped integrating with it, matrix was never good but it's the most modern option, and the atproto based ones might be good but it isn't a messaging protocol

"oh you'd have data portability" who cares? "you'd have federated identity!" what is that? that sounds like a nerd thing "big companies don't control your data" hate to say it but the average person will not give up something because of privacy, even if they know it's bad

kinda sad that there probably won't be a good federated chat protocol again. irc was good for the time, xmpp was killed because big companies stopped integrating with it, matrix was never good but it's the most modern option, and the atproto based ones might be good but it isn't a messaging protocol

if you care about the crease in a folding phone's screen, you honestly should just not get a folding phone. it's impossible to get rid of it and anyone telling you different is trying to sell you their phone

i kind of wonder how possible it is to fine-tune an either transcription model or very small OLM that does cleanup only, like S1-mini, and to make that fine-tuned on all of your typing styles for various things (like emailing coworkers and messaging friends as two separate models)

oh fuck, you could totally do some kind of sync thing type thing where you move the model between devices oh fuck, you could even do parametric memory systems too... I don't know if that's really applicable to a transcription model, but it'd be funny if you could do that

i kind of wonder how possible it is to fine-tune an either transcription model or very small OLM that does cleanup only, like S1-mini, and to make that fine-tuned on all of your typing styles for various things (like emailing coworkers and messaging friends as two separate models)

I should vibecode my own transcription software and AI client for my Operating System Of The Future... aka i kinda just want a hotkey where i can point a model at an image and get good alt text for it but idk if there exists a purpose built model for that

oh also a translation interface on device?? save me?? maybe one that's transparently able to do OCR into the translation model? maybe i'm able to actually use the NPU on modern computers? maybe it's not just "we have an llm for all of this smile"? many tiny models for tasks instead of a big one?

i really wish that you could not own a smart tv in today's world, but the only way you can do this is by getting a tv that's designed for commercial services and 24/7 operation with a shit picture...

. o O ( IoT vlan that doesn't allow other devices to communicate to each other or the internet, but every wifi connected IoT device has tailscale on it and uses that to communicate back to the home assistant server only over specific ports...) firewall rules? what's that?

ok, damn

sorry 160 tok/s is way more worth it than having slightly more intelligence at 40 tok/s (on fast mode, not on it it's like 20-30 lol)