You’re reading the 24th marimo newsletter.
marimo team at PyCon. The marimo team will have a large contingent at PyCon US from May 15-17. Come hang out with us! We’ll set up a #pycon-us-2026 channel on Discord for communication.
Learn to use marimo for AI and ML development. We’re hosting a free live workshop with our friends at O’Reilly: marimo for AI & ML Development. Parul Pandey, Kaggle Grandmaster and author of Machine Learning for High Risk Applications, is teaching a hands-on session covering both foundational and advanced concepts for using marimo in ML and AI workflows.
Announcing our notebook competition winners!
We wrapped our first ever notebook competition with alphaXiv, in which participants were asked to implement AI research papers as interactive marimo notebooks. We had 143 different submissions across 44 different papers, with our top 3 utilized papers being The Geometry of Noise, Exclusive Self Attention, and Training Language Models via Neural Cellular Automata.
We had so many wonderful submissions. Without further ado, our winners:
First place. Akash Sharma - Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights
Second place. Jesse Hartman - The Dead Salmons of AI Interpretability
Third place. Konstantin Taletskiy - Geometry Of Noise
All the winning notebooks have been added to the community gallery, and some others, too! They’ve also been added to alphaXiv as associated resources. Thank you to all who participated, all who cheered on our community, and all who came to our event yesterday! And if you enjoyed this one or this made you want to participate…you’re going to love our next one 😉.
Connect agents to marimo sandboxes on molab with marimo pair
Last month we launched marimo pair, a skill that drops agents inside a running marimo notebook session. Agents can do anything humans can do in the notebook and more, including reading variables, editing and running cells, and even manipulating UI elements. It turns marimo into a collaborative canvas for humans and agents, giving models access to a far richer state and action space than they have when working with scripts — as one of our users put it, “marimo pair is the best invention of the 21st century” 🍃.
We’re excited to announce that marimo pair is now fully integrated with molab, giving your agents an isolated online sandbox for research and data work. Check out our video explainer to learn how to get started!
molab is evolving …
molab is the easiest way to run marimo notebooks in the cloud: you can create new notebooks in isolated sandboxes, preview notebooks hosted on GitHub, fork others’ notebooks, share notebooks as apps, and even embed interactive notebooks in your own docs and websites.
Behind the scenes we’ve been hard at work migrating marimo to run on CoreWeave’s infrastructure, which will allow us to provide far more computational resources (including GPUs) and services to our community (you!). We’ll have more to share soon, but in the meantime we’re rolling out access as part of a private preview. For our preview, we’re looking for users and partners interested in building notebooks related to AI/ML; if that’s you, reply to this email and let us know!
P.S. To get inspired, check out our community gallery; share your notebook on socials for a chance to be featured!
Open-source updates
We’re shipping as fast as ever. We have two exciting new features to share.
Slides minimap
We’re giving our slides feature a makeover. We’re migrating to RevealJS and making them a whole lot more customizable. We’re releasing new features incrementally; to start, we’ve added a minimap that makes it easy to navigate your deck while you work on it. Stay tuned for many more updates.
Table glow-up
Tables are also getting a glow-up. We’ve streamlined the user interface in many small but delightful ways. Most recently, we added an editable filter UI.
This new UI lifts filters (such as numerical column between two values, or string column contains substring) into editable pills, letting you update filters with just a couple clicks and reducing the distance between you and your data.
🍃 Community
We’re at 20k+ GitHub stars, 230+ contributors, and 3.6k+ members in Discord — join the conversation!
Here are some mentions of marimo in the wild that we found particularly inspiring.
- Operational health metrics dashboard. dltHub checks freshness of data for dashboards, alerts, or ML models.
- Polynomials Argument Experiments. Simone Conradi, Ph.D. plays with polynomial p(z) of degree 100, z ∈ ℂ, with each of the 100 roots drifting along its own circular orbit.
- Matrix factorization. Pragyan Tiwari explores Orthonormal Basis with Gram-Schmidt.
- ML Research. Julian Dralle shares his master’s thesis, training a discrete representation model end-to-end.
- Multimodal Cancer Research. Aignostics shares a production suite of marimo notebooks for spatial tumor microenvironment analysis.
- Astronomy. Austen Wallis gives talks on astrophysical applications, comparisons of Neural Network vs Gaussian Process designs for physical emulation, and deployment of accessible, real-time diagnostic tools via marimo notebooks.
Don’t forget to check out our curated community gallery. Share your molab notebooks on social media and tag us for a chance to be featured!
Sincerely,
marimo team 🍃
