Play real scenarios
Train models, fix data, and make decisions inside vivid missions.
Let's get started!
Hello, learner ✦
YOUR NEXT LITTLE ADVENTURE
Teach a robot to
spot a cat.
See how a neural network finds patterns — one tiny decision at a time.
◷ 8 min
HANDS-ON LAB
Make the machine guess
Would this get a treat?
Adjust what our tiny model knows about an animal. Watch its confidence change.
76%
MODEL'S GUESS
Probably a cat
The sleepy, pointy-eared clues are convincing.THE MAIN QUEST · KNOWLEDGE CHECK
Neon City: AI Missions
Walk through a living future city. Each district gives you a real AI scenario to solve. Apply what you know to unlock the next mission.
0 / 5 missions◈ Knowledge rank: Rookie
LEARN THROUGH VISUALS
Watch a model learn
From examples to a smart guess
A model studies labeled examples, notices patterns, then uses those patterns to make a new prediction.
🐱🐶Labeled examples
→
✦Finds patterns
→
🐱92% cat
“Pointy ears” and “whiskers” are clues the model can learn from.
ARTICLE MODE · 6 MIN READ
How a model learns to spot a cat
Supervised learning is simply learning from examples that already have the right answer attached.
Start with labeled examples
Imagine a folder of photos. Each cat image has a cat label, and each dog image has a dog label. The label is the teaching signal: it tells the model what it should learn to predict.
photo_001.jpg → cat
photo_002.jpg → dog
photo_003.jpg → catLook for useful patterns
The model turns each image into measurable clues called features. It may notice whiskers, ear shape, fur texture, and the outline of a face. No single clue has to decide the answer alone.
features = [whiskers, ear_shape, fur_texture]
model.fit(features, labels)Make a new prediction
When a new image arrives, the model compares its clues with patterns it learned before. Its output is a probability, not a feeling of certainty.
prediction = model.predict(new_photo)
# { cat: 0.92, dog: 0.08 }Why labels matter
Correct, varied labels help a model learn. Unlabeled or incorrect photos make the training signal weaker, which is why the Neon City mission rewards you for taking labeled data to the model server.
You’ve reached the end. Try the visual mode again to see the same idea in motion. ✦