A notebook is made of cells. A code cell might load a spreadsheet of customer orders; run it, and the table appears right below. The next cell draws a chart of monthly sales, and a text cell explains what the chart shows. Because each step can be run, changed and rerun on its own, notebooks suit exploration: poking at a dataset, trying a model on a few examples, or teaching someone how a technique works. Python is the usual language, though Jupyter supports many others.
The format is open, so notebooks run in many places: JupyterLab on a laptop, Visual Studio Code, or hosted services such as Google Colab that come with free or paid GPU and TPU time. That is why model tutorials, from PyTorch guides to Hugging Face courses, are often shared as notebooks.
The same freedom causes problems in production. Cells can be run out of order, so a notebook may only work because of a step someone ran earlier and later deleted. Saved outputs can include sensitive data, and the files are awkward to review in version control. Teams usually move code that works in a notebook into tested scripts or pipelines before relying on it.
In episode 47 at 26:38, Google DeepMind’s Paige Bailey demos Colab, “a Python environment where you can write code in a notebook or pull up a terminal,” with Gemini CLI built in; she works on Google’s developer tools, so it is a demo of her company’s product. Episode 47 was sponsored by Galileo (sponsor thanks at 49:12).
Hear it from the guest
“It's a Python environment where you can write code in a notebook or pull up a terminal.”
Quotes lightly edited to remove filler words.
Sources
- Project Jupyter documentation — Describes Jupyter as an umbrella project, including the Jupyter Notebook and JupyterLab web-based notebook applications, centered on tools and standards for interactive computing; checked October 24, 2026.
- nbformat: The Jupyter Notebook Format — Official documentation of the open file format (.ipynb) that stores a notebook's code, text and outputs; checked October 24, 2026.
- Google Colab FAQ — Describes Colab as a hosted Jupyter Notebook service that needs no setup and provides free access to computing resources including GPUs and TPUs, suited to machine learning, data science and education; checked October 24, 2026.
