Let a reader run the notebook without installing anything

[DenisDrobyshev]

Every module linked its notebook to a GitHub blob page. That renders the code and will not run it. To try anything, a reader had to clone the repository, install Python, install the requirements and start Jupyter — a barrier in front of all twenty-seven modules of a course whose argument is that you should check a claim yourself rather than believe it.

Why this is free

The notebooks import NumPy and Matplotlib and nothing else — the rest is the standard library, checked across all 27:

numpy 27, matplotlib 27, csv 1, collections 1, math 1, re 1

Both are already in Colab. So the link costs nothing to offer and nothing to install: click, and it runs. prerequisites.md has been recommending Colab for parts III–VI all along; there was just no way to get there.

The order matters

The Colab link comes first, the filename second:

Ноутбук: открыть в Colab · notebooks/01-claim-baseline-noise.ipynb

Rendered the other way round it was the last item on a dense metadata line and wrapped onto a line of its own — the least prominent thing on the page, while being the most actionable. I checked this in the built site rather than guessing.

Keeping it true

scripts/check_modules.py (standard library only) requires that every module page carries both links in that shape, that every referenced notebook exists, and that every notebook is reachable from some page.

None of that is covered today: mkdocs --strict does not check outbound links, and the notebook job proves the files execute, not that a reader can find them. Verified the check fails on each case — a missing Colab link, a page pointing at a notebook that is not there — and passes on the tree as it stands.

Verified

• 54 pages updated (27 modules × 2 languages), all pass the new check
• mkdocs build --strict clean
• All 27 notebooks fetchable at the URL Colab reads from — raw.githubusercontent.com returns 200 for every one
• The Colab URL shape resolves