marimoFrom the builder

Python notebooks as dataflow graphs: reactive, reproducible, and reusable

Adapted from a PyCon 2025 talk, the post highlights notebook reproducibility problems, citing a 2019 study in which only 24% of nearly 1 million GitHub Jupyter notebooks with valid execution orders could be rerun and just 4% gave the same results. marimo statically analyzes code to build each notebook as a dataflow graph, which lets the same file act as a reactive notebook, a runnable script, a Python module and an interactive web app. The post says marimo sees hundreds of thousands of downloads a month and has more than 15k GitHub stars.

marimo is a new kind of open-source Python notebook.

Read the original

Saved . A shortened excerpt from the official blog; the original publisher retains authorship.

Show all