Physics Simulations Are What Vibe Coding Should Have Been All Along
A small open-source repo turns physics into interactive demos, and quietly shows why AI-assisted coding is so useful for learning.
There is a particularly noble kind of software project that does not try to raise a seed round, reinvent productivity, or explain why your calendar needs a blockchain.
It simply shows you something.
A magnetic field.
A planetary orbit.
A curved spacetime surface.
A solenoid doing its electromagnetic party trick.
That is why I like this little open-source project: physics-sims.
It is a collection of interactive physics simulations generated through LLM prompting. The current repo is HTML-based rather than Python/Pygame, which is actually quite sensible: clone it, open the files through Live Server, and you have browser-native educational demos without having to summon a dependency dragon from the command line.
The demos cover the sort of physics concepts that normally get murdered by static diagrams in textbooks.
There is Earth’s magnetic field, shown as a dipole tilted relative to the spin axis. There is a classic electromagnetism solenoid demo. There is a general relativity simulation for spacetime curvature. There is a planetary orbit demo with Hohmann transfer orbit logic.
In other words, actual educational material. Not another AI wrapper around a spreadsheet.
And this is where vibe coding becomes genuinely useful.
The fashionable version of vibe coding is often presented as: “I typed one sentence, the model built an app, and now I am basically a software company.” Which is charming, in the same way a toddler wearing sunglasses is charming.
But the serious version is much better.
You have an idea. You understand the subject well enough to know what the simulation should show. You ask the model to build it. You test it. You look at what is wrong. You ask it to fix the visual scale, the labels, the motion, the stability, the interaction, the edge cases. You keep going until the thing teaches.
That is not magic.
That is a new kind of workshop.
Physics is especially well suited to this because so much of it is visual and dynamic. Newton’s laws are not just equations. Gravity is not just a paragraph. Electromagnetism is not a decorative horror of arrows. These are systems. They move. They evolve. They behave.
A simulation lets the learner poke the system and watch it answer back.
That feedback loop is the point.
This is also why AI-assisted educational coding is more interesting than yet another autogenerated landing page. A landing page sells. A simulation explains. And if we can now create small, interactive explanations cheaply, then the bottleneck in education shifts.
It is no longer “can someone code the demo?”
It becomes “does someone understand the concept clearly enough to design a good demo?”
That is a much better bottleneck.
The future of learning should not be endless PDFs, static slides, and diagrams that look as if they were last updated during the reign of fax machines. It should be interactive, inspectable, remixable, and small enough for one curious person to build.
This repo is not some grand unified platform for physics education.
It is better than that.
It is a reminder that with modern AI coding tools, one person can turn an idea into a working educational object quickly enough that experimentation becomes cheap.
And when experimentation becomes cheap, teaching gets more interesting.
Not because the machine replaces understanding.
Because it gives understanding something to build with.