Software Engineering Daily Reinventing the Python Notebook with Akshay Agrawal
69 snips
Mar 10, 2026 
Akshay Agrawal, a Stanford PhD and former Google Brain researcher who created Marimo, describes a next-generation reactive Python notebook. He discusses problems with traditional notebooks. He outlines Marimo's reactive execution, static analysis, reproducibility as plain Python files, interactive UI widgets, serving notebooks as web apps, and integrations with LLMs and agents.
AI Snips
Chapters
Transcript
Episode notes
Snapshot Outputs For Sharing And Faster Loads
- If you still need embedded outputs for sharing, enable Marimo's option to snapshot an IPython notebook alongside the Python file.
- Marimo also keeps a session cache to preserve prior outputs for faster load without rerunning everything.
One File Bridges Prototyping And Production
- Because Marimo notebooks are valid Python, you can import functions or classes from them and run them as scripts or cron jobs.
- That blurs prototyping and production: prototype interactively, then run python my_notebook.py as a pipeline.
Use mo.ui Widgets To Make Variables Interactive
- Add UI widgets to variables with mo.ui so scrubbing inputs auto-runs dependent cells without callbacks.
- Each widget exposes a value attribute to read back into Python, enabling immediate reactive interactivity.
