My ambition of making inference and compute infrastructure more robust, optimized, reliable, and efficient is now backed by @join_ef. Iโll be joining the EF Fall batch.
Iโm also stepping away from college to go all-in on this.
Grateful to @rahulsamat@vatsaldusad@vaishnavireddys for taking the bet on me before the outcome is obvious.
@matthewclifford@Alicebentinck, youโre going to look back at this as one of the most asymmetric bets youโve ever made in inference and compute infrastructure.
The best part about processing 100M tokens wasnโt the 100M.
It was watching Tensormux stay reliable as usage scaled.
Thank you to everyone building with Tensormux. Weโre just getting started.
You can still claim your free api at https://t.co/QJ1hnl33wv
whatโs your GitHub character? ๐
I built a tool that analyzes your GitHub and tells you which character you code like.
mine gave me Wonder Women and then roasted my entire GitHub ๐ญ
try yours : https://t.co/hIHQXawcwY
running on @Tensormux
Your coding agent has a new free model.
@cohere's North Mini Code 30B is now live on Tensormux.
Create an API key, plug it into your workflow, and start building.
$5 free credit. 1 million tokens. No card required.
https://t.co/wWUSKBolCA
We heard you.
You wanted bigger, better models, so we're burning a little more GPU to make it happen. Now live on Tensormux Shared serving: GLM 4.7, Gemma 4, Qwen3.6 and gpt-oss. Still free, still no card.
New here? Claim up to 1,000,000 FREE TOKENS!!.
only at https://t.co/eMW7UXe8dW
We're opening Tensormux to everyone, for free, and burning our own GPUs to do it at https://t.co/eMW7UXe8dW.
Grab an API key, no credit card, and get $5 of free credit which is ONE MILLION TOKENS!! for free to run some of the best open models on the planet: Llama 3.1, Qwen3, gpt-oss and Gemma. Use it in your app, your agent, your weekend hack, wherever you want.
We built Tensormux because inference in production is supposed to be boring. Fast, reliable, and never falling over at the worst possible moment. Most people never get to feel that without a sales call and a contract.
So we're skipping all of that. Get your key. Go build something, and try your best to break it.
Kernel optimization is obsolete. Just npm install it.
We built kernel-skills at Tensormux to help our agents write, debug and optimize CUDA, Triton and quantized kernels. After months of using it to speed up inference, weโre open-sourcing it.
npm install @ krxgu/kernel-skills