An open-source AI-native Python notebook: reactive, git-friendly, execute as scripts, share as apps đ github.com/marimo-team/marim⌠đŹ marimo.io/discord
- Tweets 1,306
- Following 9
- Followers 7,088
- Likes 1,601
Pinned Tweet
marimo brings your data to life!
Select points in a plot and get them back in Python, instantly, with reactive charts.
In this data app, a reactive chart lets the user interactively explore the original images backing a 2D embedding of MNIST.
10
112
671
233,395
marimo retweeted
I spent some time exploring Causilo, a tabular foundation model from Nums AI.
There have been quite a few new models entering the tabular foundation model space recently, and Causilo is particularly interesting given its strong results on tabular benchmarks, especially considering itâs the first model from the team.
I created notebooks in both Kaggle and molab ( powerful GPUs, more CPU and RAM) to try Causilo out and explore what is available in the API, since the technical report is still awaited.
Kaggle Notebook: kaggle.com/code/parulpandey/âŚ
molab (@marimo_io notebook) : molab.marimo.io/notebooks/nbâŚ
5
16
799
We were thrilled to get early access to @typesafeai's Jev model and explore tons of new use cases!
Here's what @ktaletsk is building with Jev:
1. jevframe - pandas and polars dataframe adapter to bulk generate new columns with Jev github.com/ktaletsk/jevframe
Try the notebook: molab.marimo.io/github/ktaleâŚ
6
18
111
5,983
2. Adaptive UIs.
Jev really shines in choosing UI elements depending on the context. Interactive marimo-pets widget knows which cell you're viewing and ranks the most useful tools for that code cell.
Try here: github.com/ktaletsk/marimo-pâŚ
8
475
marimo retweeted
đ Announcing Zax: A new tensor-native data-proximate compute service from Earthmover!
Zax's query planner creates an execution plan that innately understands scientific data. Zax beta services, Zax-SQL and Zax-OpenEO, are now available to customers and free tier users. đ
2
5
10
1,373
Congratulations to our @CoreWeave Hacks "Best Use of marimo" winners, Standard Physics!
Their prototype uses iPhone scans to help shop owners spot potential accessibility issues.
Built by Brendan Giang, @IImzihao (Jerry Yan), @b_nezlobin (Boris Nezlobin) & David Wu.
ALT The four members of Standard Physics and marimo judge Konstantin Taletskiy pose together at CoreWeave Hacks after the team won Best Use of marimo.
2
1
15
907
In âOne shop, one knob at a time,â sliders reshape a shop and rerun the appâs actual accessibility checks.
That link between controls and computation made it stand out.
Notebook source (local setup required): github.com/Imhaohao/standardâŚ
ALT Standard Physicsâs marimo notebook, âOne shop, one knob at a time.â Controls adjust the gap between display cases, counter height, doorway width, and seating. Below, accessibility findings flag a counter that is too high and a route that is too narrow.
1
6
317
Thrilled to be partnering with @ValsAI on their RSI Index: a benchmark for whether models can do the work of an AI researcher.
How close are we to true RSI?
Results below, plus the full experiment ledgers in marimo notebooks so you can watch each model try to build its own successor âŹď¸
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.
1
6
19
2,003
Awesome work from our friends at @askalphaxiv: they used Tinker to give agents a fixed, cost-predictable budget, then had the agents reproduce SDFT's continual learning gains, proving self-distillation claims can actually be checked
Check out the linked marimo notebook!
Using Tinker with an autoresearch loop is a really effective way to reproduce post-training papers at predictable costs
Today there are dozens of self-distillation methods all claiming improvements over each other, and itâs hard to establish which claims hold up
We gave agents a Tinker budget to reproduce self-distillation results across models and training setups. With just a few user prompts, they reproduced SDFTâs continual learning benefits across Qwen3-8B and Qwen3-30B-A3B over multiple seeds, and investigated SFTâs failure modes.
Read more below:
1
1
9
1,091
For our cheminformatics competition, you can now win an @Apple Mac Mini M6. You have 1 month left to take a dataset and build a marimo notebook that brings cheminformatics to life. Competition details below.
1
6
24
2,822
Competition details:
marimo.io/pages/events/notebâŚ
1
3
381
33 days to win a mac mini.
the challenge: pick a cheminformatics dataset and build a marimo notebook that answers a question you want the answer to.
we're judging based on creativity, not on model accuracy. co-hosted with @wpwalters.
we'll see you in our submissions đ
1
4
44
4,031
