From video calls to a course you can do on your own

In late June 2026 a friend asked whether we could help someone learn to build websites: HTML, CSS, JavaScript and React. The student was an adult who already lectures in another technology field, robotics, and wanted to start teaching full stack. We said yes and began the way most people would, with a video call.

Key points

• We gave twelve live lessons on video calls, between 23 June and 7 July 2026.
• The live format did not fit, so we decided to build a course that needs nobody watching: read, learn, practise, then the next lesson.
• From 12 July we stopped teaching and started building.
• A question came up in the live lessons: why solve it yourself when an AI can? Our answer is the difference between having it write the code and asking it to explain.
• You can step away and come back: your draft is saved as you type, and every result is explained in plain words.

What did not fit

A live lesson asks two people to move at the same pace, at the same moment. Learning to code does not work like that. You read a question, try something, get stuck, read the question again, try again. Some of that wants quiet. Some of it wants repeating. On a call, the teacher ends up saying the same thing twice, and the student waits for the next thing to be said.

Nobody was doing anything wrong; a live call is the wrong format for this kind of learning. After twelve lessons we agreed on a different plan: we would build something the student could use alone.

The idea: read, learn, practise, next

The plan was four steps that repeat. Read a short lesson. Try an exercise. Get an answer from the computer. Move on to the next lesson. Nobody has to watch, and nobody has to repeat themselves.

1. 23 June 2026The first live lesson on a video call
2. 7 July 2026The last of twelve live lessons
3. 12 July 2026We stopped teaching and started building
4. 13 July 2026The first version of the course was saved in the project history

Built so you can leave and come back

Most of what we added is small, but each piece makes it easier to leave and come back.

• You press Run Tests and get a list of checks, each one in plain words, for example that a paragraph has a certain style. If an answer is wrong, you can set it beside the question and read the question again.
• There is a Hint button and a Solution button, and the results can be read aloud (for Hebrew, this works only if a Hebrew voice is installed on the device and the browser supports it).
• Your draft is saved as you type, and your solved lessons are kept in the browser on that device. Close the tab and come back later, and you are where you were.
• The lessons that ask a short question have a button that says “I don’t understand, go back to the previous lesson”.

Why solve it yourself, when an assistant can?

The question came up in the live lessons, and it comes up everywhere now: “Why should I solve it? Let the AI do it.” We explained to the student that there are two ways of working with an AI, and they lead to very different results. One is asking an assistant to write the code. You get something that runs, and no idea why. The other is asking it to explain: what does this line do, why did my answer fail, say it again in simpler words.

We cannot stop anyone from pasting an exercise into an assistant and pasting the answer back. The course’s checks will say yes, and nothing will have been learned. What we can do is lean the other way: each lesson is short, so the student’s own answer is the point, and the short questions that end many chapters check that you understood the idea, not only that the code runs.

An assistant is often very good at the second. It can explain a point in minutes, in words chosen for whoever is asking, and a beginner, a student and an experienced engineer each need different words. So the course keeps the solving with the student, where understanding is built, and it teaches how to use an assistant for explanations. One of its lessons, “Working with an AI Assistant”, says it plainly: ask for an explanation, not just code, and treat the assistant as a partner, not a replacement for understanding.

Where the idea came from

While we were planning, we found JS Hero (opens in a new tab), a free JavaScript course from Germany that checks your answers in the browser. Seeing that it could be done was the spark. We want to thank them for the idea. What we built from there is our own, and the next stories explain how: how the answers are checked, how the lessons were written and the code editor in the box.

What we learned

Teaching one person live showed us what the course needed to be: something that never depends on someone being there at the right moment. It was built so you can leave it and come back to it.