Karanbir Singh ๐ง
4.8K posts
Karanbir Singh ๐ง
@KaranOrg
RedHat; Reliability & Scale Engineering; {manage, automate, secure} everything. ( ex-CentOS project lead )
London & Cambridge, UK ๊ฐ์
์ผ Kasฤฑm 2008
99 ํ๋ก์15K ํ๋ก์
Karanbir Singh ๐ง ๋ฆฌํธ์ํจ
Launch your private #AI code assistant today. The new AI quickstart, powered by @RedHat_AI's MaaS and NVIDIA Nemotron models, gives you governed, cost-controlled AI. Learn how: red.ht/4uwrQ9r #NVIDIAGTC
English
Karanbir Singh ๐ง ๋ฆฌํธ์ํจ
๐ข Applications for Google Summer of Code (GSoC) are open now through March 31, 2026. Spend 12+ weeks coding with experienced mentors and earn a stipend.
Read more at goo.gle/gsoc-2026-applโฆ
#GSoC #OpenSource #GoogleSummerofCode #StudentPrograms
English
Handy list of peering exchanges around pulse.internetsociety.org/en/ixp-trackerโฆ ( you can select the country and get local details as well )
English
London internet exchange flow traffic graphs portal.linx.net/services/lans-โฆ
English
I don't think you needed to look above the rack, what's going on inside the rack isn't all that great either.
RossRadio@cqcqcqdx
Contractors have taken down the roof above a cabinet, showing the gore installed by the previous IT team...
English
What's the worst that could happen ? Go on, unleash the AI on all your remote machines, in real time!
( For when used responsibly, this is quite cool tbh )
Anthony Morris ใ@amorriscode
SSH support is now available for Claude Code on desktop Connect to your remote machines and let Claude cook, TMUX optional.
English
@FFmpeg Other than $$ and Development Time, is there any other meaningful contribution you would find useful?
English
Everyone with the DGX Spark ( or any of the GB10 machines ) may want to look at this..
Red Hat AI@RedHat_AI
Red Hat AI just released NVFP4-quantized models from 8B to 400B+ parameters, built with LLM Compressor and ready for vLLM deployment. Explore the collection: ๐ huggingface.co/collections/Reโฆ Check out the ๐งต for details on accuracy recovery and more!
English
Karanbir Singh ๐ง ๋ฆฌํธ์ํจ
vLLM got a major upgrade!
Originally built for autoregressive text-based LLM serving, vLLM-Omni now lets you serve text, image, video, and audio models - all from a single framework.
You can also serve diffusion models for fast parallel generation.
100% open-source.
English
I know this is meant to be a bit of a retro throwback. But I know my people using trolleys in the Data Centre still ticking the box, TFT screens and maybe no VCR but trolley KVM carts are a daily thing.
World of Engineering@engineers_feed
Name something you remember watching on this:
English
Spare a thought for all the Linux 2.6.x out there ( yes, there is a lot of it )
@mikko@mikko
@isurusam YES.
English
Karanbir Singh ๐ง ๋ฆฌํธ์ํจ
The CIA just stopped publishing their World Factbook and took every page, including the archived copies of previous versions!
This sucks. It was public domain, so I recovered the 2020 edition (the last one published as a zip file) and shared it to GitHub simonwillison.net/2026/Feb/5/theโฆ
English
English
@FFmpeg @jungpionier_ Not only are they available, but they're fast... 5GHz, 32-core/256-thread ... Onboard accelerators ... True SMP scalability ...
It's a real shame SPARC got put out to pasture. The M8 chip was a beast and only now is the x86 space catching up.
(M8 launched in 2017.)
English
This would be a great machine to revive FFmpeg's Solaris CI
Adam Rossi@rossiadam
In 2004 the Sun Fire E25k was the apex predator of computing. This was the largest single-image Unix machine that money could buy ($2M fully loaded). This bad boy had: โข 36 UltraSPARC IV processors (72 cores) โข 576 GB RAM โข 18 fully hot swappable CPU/memory boards โข gigabit ethernet and fibre channel HBAs โข Redundant power supplies, cooling zones, draw 15kW of power I deployed software to these things. Uptime was measured in years. This was THE ONE. Man I miss Sun.
English