marimo

marimo is a next-generation Python notebook that fixes the long-standing Jupyter pain points: hidden state, out-of-order execution, broken Git diffs, and the gap between a notebook and a deployable app. Cells in marimo form a reactive dependency graph — change a variable and downstream cells re-run automatically — which makes notebooks behave more like spreadsheets and eliminates the classic 'I forgot to run cell 5' bug. Notebooks are stored as pure .py files, so they diff cleanly in Git and can be imported as ordinary Python modules. With a single command (marimo run) you can serve any notebook as an interactive web app with sliders, dropdowns, tables, and plots — useful for AI demos, model dashboards, internal tools, and teaching materials. marimo includes a built-in AI assistant that can generate or edit cells using your choice of model (OpenAI, Anthropic, local Ollama). It also ships first-class support for SQL cells, DuckDB, and Polars, which makes it a strong choice for data-heavy AI workflows. The project is permissively open source (Apache 2.0) and is rapidly gaining traction inside ML teams as a Jupyter replacement.