I saw Ryan Baker’s video, How the Shell Works, and it reminded me of a talk Derek Muller from Veritasium gave, What Everyone Gets Wrong About AI and Learning. It got me thinking about the future of teaching someone these concepts. How much of a video like Ryan’s could be generated with code, and how much still needs someone carefully curating it with a learner in mind?

So I made a small proof of concept. One concept on the screen and a voice over explaining it. The animation is written in Python with Manim Community Edition, the engine Grant Sanderson created for 3Blue1Brown. The narration comes from ElevenLabs, which I had heard about but never used until now. The scene walks through the difference between the terminal emulator, the shell process, and the kernel, then shows $SHELL being looked up in the environment and printed.

I found it to be an interesting alternative to a PowerPoint. The code describes both what moves on the screen and when the voice says it. If you were presenting it, you could pause at any moment to take questions or whiteboard something.

The fiddly parts were pacing the animations to match how fast a person actually follows along, and rewording lines so the voice did not read slashes out loud. “/bin/zsh” became “the path to the Z shell binary.” The same script can render with ElevenLabs, with free gTTS, or silent for quick CI runs. I also used it as a chance to try the agy agent CLI as the main builder, pointing it at my Python project template so it could bring over the CI, Makefile, and Sphinx docs setup.

Try It

Source is on GitHub.