A playground for mathematical curiosity.
See the idea.
Understand the math.
Some things click when you can move them.
Explore a concept, change a parameter, follow what happens. Then make a playground for your own questions.
Show demoFrom “what if” to “oh, I see.”
Your question.
Your experiment.
The examples are a starting point. The local app lets you build something around the idea you’re trying to understand.
- 1
Bring a question.
Describe a concept, paste a problem, or attach your notes. Use the Codex CLI signed in on your computer.
- 2
Make it tangible.
Get an interactive visualization with plots, controls and numerical readouts. Change a parameter and see the effect.
- 3
Keep following the idea.
Ask for a new view or another control. Your conversation, source code and earlier versions stay together.
A little lab.
On your computer.
Clone the full app repository and start your own local playground. Make visualizations, keep your projects, and inspect the source behind every graph.
Setup instructionsA quick setup, then it’s yours.
- Install Node.js 22 or later.
Get it from nodejs.org ↗. npm is included.
- Clone the app and start.
Clone the separate
vizualearnerrepository. Inside its folder, runnpm start. The first launch installs dependencies automatically. - Connect your Codex account.
Install the CLI, then sign in:
npm install -g @openai/codex codex login - Open your playground.
Visit
http://127.0.0.1:3210. Keep the launcher terminal open while you work.
AI support: Currently works with Codex CLI. Claude Code and API-based integrations are planned.
The app runs from source, not a signed native installer. Your own Codex usage limits and provider charges apply. By default, the app makes AI calls only when you request a generation. The bundled example works without a Codex login.
A few things worth knowing.
What does the app repository include?
The complete runnable app, its source, tests, launchers and documentation. This website is a separate repository. Local projects, conversations and credentials are excluded from Git; dependencies are installed on first launch.
Is this a hosted AI service?
No. The public examples run in your browser. The local app runs on your computer and calls your installed Codex CLI for new generations. You do not need an account to try the prepared examples here.
Where do my projects go?
The app saves projects locally, in its .local/workspace folder or a directory you choose. Prompts, attachments and relevant source are sent through your configured Codex provider when you request an AI generation.
How reliable are the visualizations?
Generated mathematics can be wrong. Check equations, assumptions and numerical results independently before relying on them. The prepared examples demonstrate the interface; a working graph does not verify an AI-generated result.
Can I see and change the generated code?
Yes. Every visualization has a Source view, downloadable files and a revision history. The local app is available under the MIT license.