Jupyter Notebook

A Jupyter Notebook is a working document where explanatory text, code, and the results of running that code stack up from top to bottom in a single file. In Korean, "notebook" often brings to mind a laptop computer first, but when developers and researchers talk about a notebook, they mean this document, not a machine. "I opened the notebook and ran the model" doesn't mean someone opened up a laptop.

The comparison to a lab notebook is apt. A scientist's lab notebook records, in order, what they tried to do, how they did it, and what results came out. A Jupyter Notebook is a lab notebook where that middle part actually runs. The document is made up of a series of cells, alternating between cells for writing explanations and cells for writing code. Run a code cell, and a table or a chart lands right below it, saved exactly as it appears.

The name comes from the open-source Jupyter project, which started in 2014, and today it functions less like the name of a specific tool and more like the name of a format. Google Colab, Kaggle, VS Code, and Amazon SageMaker all use the same format. That's why most machine learning tutorials are distributed as notebook files — download one, click through the cells from the top, and someone else's experiment reproduces itself on your own screen.

The term is coming up again in AI agent news because the one filling in those cells is shifting from human to AI. Write a single line stating a goal in the top cell, and an agent fills in the cells below; by the time the notebook is finished, it holds a full record of what was tried, where it got stuck, and why a particular approach was chosen. Work that used to scatter and disappear gets captured, whole, in a single document.

"An engineer writes a goal in a notebook cell, and Codex fills in that notebook" — this isn't about opening and closing a laptop; it means AI is filling in a working document where code and execution results accumulate together.

1. Go to colab.research.google.com and sign in with a Google account to create a new notebook. There's nothing to install — it opens right in your browser.
2. Type print("hello") into a cell and click the run button on the left. The result appears right below that cell — that single action is essentially what a notebook is.
3. Use "+ Text" at the top to add another cell and write down "what I'm trying to do." The moment explanations, code, and results all live together in one document, you've got the notebook the articles are talking about.
4. Ask a chatbot: "Explain the difference between a Jupyter Notebook and a plain Python file using a lab notebook analogy." That will lay out why researchers bother with this format in the first place.