Artificial Intelligence · Available Now

FuzzNet Labs

Build an analogue AI model. No power required. Race your opponents to make the right prediction first. And try not to overfit your way into last place.

Open the Black Box

See inside the neural network architecture powering modern AI

AI is a Probability Machine

Why 100% accuracy is impossible and every model navigates tradeoffs

How AI Gets Built

The Design-Train-Test cycle and why data is destiny

You are AI-Fluent

Ask the right questions and push back when "the AI said so" isn't good enough

▶ Play Now. Free Online Buy the Physical Game
2–4 Players
30 Min
Ages 14+

Train it, test it, understand it.

Every player manages a neural network model. Choosing architecture, curating training data, and running test rounds against shared animal datasets. You're not just playing a game; you're experiencing exactly how AI researchers think, fail, and iterate.

What you'll actually understand after playing
  • Open the Black Box

    AI stops being mysterious when you can see inside it. Neural networks process information through layers. Input, hidden, and output. And that layered architecture is the foundation of the current AI revolution. It's the same structure powering large language models like ChatGPT. Playing FuzzNet Labs lets you visualize how data flows through those layers and why the architecture decisions you make change what your model can and can't learn.

  • AI is a Probability Machine

    No AI model ever achieves 100% accuracy. And that's not a bug, it's the nature of the problem. When you push for maximum accuracy on training data, you overfit: the model memorizes patterns instead of learning them, and it breaks on anything new. Optimize differently and you trade one kind of error for another. More false positives, fewer false negatives, or the reverse. Every AI deployment in the real world is navigating that exact tradeoff.

  • How AI Gets Built

    AI isn't conjured. It's engineered through a repeating cycle: Design the model, Train it on data, Test it against reality, then go back and adjust. Data is destiny in that cycle; what you feed the model determines everything it can possibly learn. A brilliant architecture trained on bad or incomplete data will fail. Understanding that Design-Train-Test cycle is what separates people who can work effectively with AI teams from people who are just hoping the model figures it out.

  • You are AI-Fluent

    The goal isn't to make you an AI engineer. It's to make you someone who can work with one effectively. Understanding the architecture, the tradeoffs, and the development cycle means you can ask the right questions, push back when something doesn't add up, and understand why a model behaves the way it does. Instead of accepting "the AI said so" as an answer, you'll have the mental model to dig in and demand a real one.