Marimo: A Notebook that "Compiles" Python for Reproducibility and Reusability

This conference talk from PyCon US explores the concept of creating a new type of Python notebook that understands cell execution order, updates automatically when variables change, and incorporates UI elements like sliders and dropdowns. Discover how marimo, an open-source reactive notebook, implements a "compiler" for Python through static analysis to create a directed acyclic graph representing data flow between cells. Learn about this intermediate representation that powers interactive computing while eliminating hidden state, making notebooks executable as scripts and shareable as apps. Examine the broader question of compiling Python code into flexible intermediate representations that preserve program semantics, including discussions on design decisions and tradeoffs between static analysis and runtime tracing. The 47-minute presentation uses marimo's internals as a case study to demonstrate approaches for building more reproducible and reusable Python notebook experiences.