Got invited to demo @clusyio at the AI Demo Series by @StartupGrind.
We trained a full 98% accuracy neural network in under 2 minutes using our AI-powered Jupyter IDE. The reactions were amazing.
Anyone can now train and fine-tune research-grade models with a prompt.
Can we predict whether a drug molecule can cross from the bloodstream into the brain?
We trained a model on Clusy to do exactly that. 🧠💊
Here’s the exact demo video @eldar_hsnv presented over at the @claudeai community event
Got invited to give a (very) rapid demo of @clusyio at the @claudeai Community for Science event.
Great to connect with scientists and fellow researchers, and even better to hear feedback that reinforced why we’re building this.
Time to accelerate domain experts!
Got invited to give a (very) rapid demo of @clusyio at the @claudeai Community for Science event.
Great to connect with scientists and fellow researchers, and even better to hear feedback that reinforced why we’re building this.
Time to accelerate domain experts!
CUDA OOM should not mean game over.
Clusy can move a PyTorch run to a bigger GPU without losing progress.
T4 → OOM → A100 → continue
Model, optimizer, tensors, variables, files, environment - all carried over.
The GPU is disposable. The run isn’t.
Mission: kill CUDA OOM errors.
We just shipped state-preserving GPU sandbox upscaling in Clusy:
T4 → CUDA OOM → A100 → continue
Your model, optimizer, tensors, variables, files, and environment survive the sandbox upgrade.
Clusy is the first notebook environment to preserve a live @PyTorch session across GPU sandboxes.
The environment may change, training continues
The frontier model price war just hit $0.
For the next 72 hours, Kimi K3, GLM 5.2, and Qwen3.8 Max are free for everyone on Clusy.
• Free accounts included.
• No subscription.
• No card.
Every turn on all three stays off your usage allowance.
Ends 11:59 PM PDT on August 8.
This is your chance to train a model you have always needed: https://t.co/DRlqdz2lFl
Kimi K3 is back on Clusy 🔥
The SOTA open-source model is live again.
Comment what you want to build with it. I’ll pick a few replies and bless each with $100 in Clusy credits.
An AI may have just helped break an 87-year-old math conjecture.
We dropped the counterexample from the guys at @AnthropicAI into @clusyio and verified it: the Jacobian is nonzero everywhere, yet 3 different inputs collapse to the same output.
Math is getting weird.
How much is a used car actually worth?
I gave Clusy one prompt. It found the data, trained and compared several models, then predicted resale prices for different cars.
A simple demo of going from an idea to a working ML model in minutes.