Sudo su

@sudoingX

GPU/local LLM. more RAM and OSS... everywhere

Bangkok, Thailand

Joined August 2022

1,158 Following

40.1K Followers

Sudo su
@sudoingX
14h
llama.cpp v1 is up now after 10,000+ builds and to me it's the most important repo in local ai if you're getting started this stack is what you should be approaching instead of ollama or lm studio or any wrapper, they all run on georgi's ggml underneath anyway, learn it from
Georgi Gerganov
@ggerganov
Nov 13
llama.cpp v1.0.0 is now available huggingface.co/blog/ggml-org/…
18
Sudo su
@sudoingX
15h
i've been heads down all week and meanwhile nous research dropped arete 27b, so tonight i opened the weights this is qwen 3.8 27b post trained on about 53 million tokens, and the mtp head didn't make it. the original ships 15 mtp tensors and arete has 0, with the mtp layers set
Nous Research
@NousResearch
Nov 6
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of
6
Sudo su
@sudoingX
16h
if manufacturing smaller transistors gets harder and more expensive every generation, does performance per dollar eventually matter more than shrinking the process node?
5
Sudo su
@sudoingX
17h
man i didn't know ai would play this big of a role in my life, it makes me money and optimizes my workflows and now it handles my life and my finances too
2
Sudo su
@sudoingX
17h
BREAKING: a day after naming its first ever chair, nvidia's board accepts jensen huang's resignation after 33 years and starts searching for someone who can explain what a dgx spark is interim ceo colette kress keeps the cfo job too, so for now she approves her own budget
5

Sudo su

Sudo su’s posts