An .ipynb file opens as cells, not JSON. Create one with New file… and a name ending in .ipynb — the lab writes a valid empty notebook, never a corrupt one.
Running cells
- ⇧↵ runs the cell and moves to the next one (adding a new cell at the end).
- ⌘↵ runs the cell and stays.
- The ▶ in a cell's gutter does the same; the bracket beside it counts executions (
[3]). - Toolbar: + Code, + Markdown, Run all (sequential — cell 3 sees what cell 2 defined), Stop (interrupts the running cell and ends a Run all there — the cells after it do not run, and the stopped cell says how many; a cell that catches the interrupt turns the button into Force stop, which restarts the kernel and loses your variables), Clear outputs, Restart (fresh interpreter, variables gone). A Stop pressed anywhere else — the tab strip's, on a
.pytab — reaches the same kernel and ends a Run all too. - The right end of the toolbar shows the kernel's state in the status bar's words —
kernel booting — Loading NumPy…while the first cell waits for Python. Stop is not offered during that boot: nothing is running yet. A cell run while the kernel is still warming up waits for it and starts when Python is up. - One cell runs at a time. While one runs, the other cells' ▶ are off and ⇧↵ is refused with a line saying why — a queued cell would run after your Stop.
Cells share one kernel and one namespace with each other — define x in a cell and read it in the next, exactly Jupyter's contract. (▶ Run file on a .py runs that file as a fresh module: it sees the project tree, not your notebook's variables.) Before each run the project tree is synced into the kernel, so a cell can import ansatz from the file next to the notebook.
Cell tools
Hover a cell for its actions on the right: ✎ edit a rendered markdown cell (or double-click it), md / py convert between markdown and code, ↑ ↓ move, + add a cell below, ✕ delete. A markdown cell renders on ⇧↵ and stays rendered until you edit it.
Outputs
Text, tracebacks, matplotlib figures and rich results render under the cell. Outputs are saved into the file with the notebook — that is what makes a shared notebook arrive with its plots, and why Clear outputs exists.
Packages
%pip install <name> in a cell installs pure-Python packages into the kernel for this session. Compiled scientific packages (qiskit, scipy, pennylane…) have no browser build; the kernel refuses them with a plain sentence rather than a resolver traceback. numpy and matplotlib are already there.
The Run rail while a notebook is open
It steps back: "Cells run from the notebook's own toolbar" and shows the Machines — code names reference so a cell that names a machine can copy the codename exactly.
Persistence
A notebook saves like any file — autosave, ⌘S, conflict handling. If a notebook's JSON is unreadable the view says so and shows the raw text under a warning bar so you can fix or replace it.
Related: Python in the browser · Sharing & GitHub