Marin's hero run is training on our @wandb, and a ten-person open lab is showing everyone how a 535B model actually gets built.
Pull up a chair to the show:
Runaway AI agent behavior can ruin user trust in seconds.
Episode 17 of AI Cloud Essentials breaks down why observability and custom evaluation heuristics are non-negotiable for production models.
Full episode:
Show more
Hot take vs. Hot Lap.
Nobody brags about utilization at a race. Lap time is the number.
In training, that number is goodput, the share of GPU time that actually moved the model. On CoreWeave, up to 96%.
Show more
SF, get ready. #FullyConnected26# is almost here.
Featuring:
• A keynote from @drfeifei
• Industry leaders from @Nasdaq, @JPMorgan, @Ford, @nvidia and more
• Showcases from @BattleBots and @AstonMartinF1
• A @pitbull concert!
🗓️ Sept 29 – Oct 1, 2026 | Moscone South, SF
Save your spot:
Show more
Ship now, fix later. Corey Sanders says that's...mostly right.
Our SVP of Product urges teams to get something out, because production is where you learn.
Follow the rules, keep it secure, then ship. You won't learn if you don't.
Catch the full episode:
Show more
From vibe code to production-grade AI agents.
Join us on 9/23 for What's New Wednesdays to learn how to trace agent behavior, eliminate blind spots, and leverage CoreWeave ARIA to turn observability data into reliable workflows.
Register here:
Show more
What does it actually take to run AI in production?
Adrien Gaidon, Co-founder of @WaldenRobotics, is sharing real-world insights from the frontlines of agent infra and frontier research at #FullyConnected26#.
Grab your pass and we'll see you on 9/29!
Show more
Running AI in production demands infrastructure purpose-built for scale.
At #FullyConnected26#, gain practical insights you can deploy immediately:
🔹 7 Hands-On Labs: build in production-realistic environments
🔹 8 Demo Stations: experience the latest CoreWeave cloud innovations
🔹 30 Breakout Sessions: dedicated tracks for infrastructure, models and agents
Join us Sept 29th to Oct 1st at Moscone South, SF
Register here:
Show more
Distillation isn't a research project anymore. It's a pipeline.
Weave traces → dataset curation → Serverless SFT → eval → deployment.
Join us for 40 minutes on turning real production usage into a smaller student model that meets your bar.
Show more
Kickoff of day 1 at @CoreWeave Hacks. 200+ builders, one mission: build an agent loop that catches its own mistakes.
@agihouse_org, @typesafeai and @marimo_io are here with $1k+ in credits and tons of compute. 24 hours to submit.
One team is going home with a robot dog!
Show more
Metal on metal. Sparks on the floor. That's right: the @BattleBots are coming to #FullyConnected26#. And these bots are the heavy weights... 250lbs and over 👀
Come for the carnage. Stay for the conversation about what physical AI takes to train.
Spots are limited 🤖
Show more
Mr. Worldwide has entered the chat 🌎 🫳 🎤 🔥
Rapper, singer, and songwriter @Pitbull will headline #FullyConnected26#. See you there.
CoreWeave Engineering might have discovered a ✨problem✨
When I talk with investors about AI infrastructure, GPUs are the starting point. Durability comes from making them useful: software, security, & expertise that help customers build and run products. That’s the business we’re building at CoreWeave. $CRWV
Show more
3 days. 3 tracks. 2,000+ AI leaders.
Fully Connected 26 is where the pioneers of super-intelligence gather to shape the future of AI. Expect hands-on labs, live demos, and keynotes from the sharpest minds in the industry.
Sign up now:
Show more
AI safety is back in the spotlight, driven by concerns about recursive self improvement. We built the first third party benchmark to measure just how close AI is building its own successor.
In collaboration with @marimo_io and @CoreWeave we built the The RSI Index, which runs every frontier model through the same AI research tasks and scores each one against the strongest published results. So far, the models can do the work, but is far from the human frontier.
Show more
Scaling up from low-concurrency demos to full production isn't a linear equation.
September 17th, join us to see how vLLM serving, scheduling, KV cache, quantization, and CoreWeave infrastructure shape predictable inference. Register:
Show more
Stop burning compute budget on frontier pricing when you don't have to.
In Episode 4 of Training Tuesdays, CoreWeave’s David Corbitt (Senior Product Manager) and Angky William (Engineering Manager) unpack what model distillation and post-training optimization actually deliver for your bottom line.
Save your spot before 9/22:
Show more
Compute was never the hard part of industrial AI.
Yes, an AI cloud just said that.
Physical AI models fail on data, not architecture.
The engineer who understands physics and the one who can build the model are almost never the same person. Ours are both embedded in your team.
CoreWeave Physical AI Field Engineering is live now:
Show more
This just in: Mr. Worldwide is coming to San Francisco 🌎🎤
Rapper, singer, and songwriter @pitbull will headline #FullyConnected26#.
See you there.
Show more
