

to me and probably to over 2B people


to me and probably to over 2B people


no other platform even compares


you can self host librechat, or use their hosted front of you don’t want to bother but still prefer to use something open source


try helium
ah missed this was android - I use Firefox mobile with ublock origin and it works well.
what I do feel is how slow poetry is for resolving dependencies
I used poetry for years before uv, but uv is just a better tool. If openai enshittifies it, I have no doubt it’ll be forked instantly.


it does, unless the reader has an OCR mode of sorts


they really cooked with that video, got me more excited than with most trailers
yes, and give uv a try
https://docs.astral.sh/uv/getting-started/installation/
uv init + project management commands uv add, uv remove, uv sync, uv run my_spaghetti.py, …
you probably don’t need pip or uv pip commands
and here’s how to use it in docker
It’s right, you shouldn’t be using requirements.txt in 2026


how many hours per day on avg? Just curious


in my area I struggle to distinguish D, O, and 0. Idk if someone was paid to design that, but they did an awful job


hey, remember when that was a concern of countries under dictatorship?
Well, it still is. The US is becoming one.


sitting at a desk is mostly fine, as soon as you start moving it’s annoying as hell


though anthropic could add /public/artifacts/* to https://claude.ai/robots.txt to prevent any discovered URLs from being included in search results; or the <meta> tag like the article suggests


I’d prefer not to have a car, but this country makes it really inconvenient to do it, so I have one. That’s not a double standard, that’s adapting to your reality. Maybe the sec team arrived to the conclusion that they wouldn’t be as effective without them in a society under the 2nd amendment.


As someone who said it before, I was recommending that to technical people having technical problems because of the stupid way windows does some things.
So while I can’t speak for anyone saying that, it doesn’t mean it’ll always work, or even that you’ll have less problems, but it does usually mean that it’s easier to figure out what’s wrong and how to fix it because the system is so open.


sorry, I’m not a compiler, so no


lmao
for inference you’re only counting active parameters towards VRAM, and some labs / models don’t train at 32b precision, or even use the same precision for different parts of the network