Three tiny games, one for each way machines learn — sort examples, find hidden groups, and win by trial and error. No login, nothing saved.
Each dot is an example you already know the answer for — green "approved", blue "declined". Drag the sliders to draw a line that separates them, like a new employee learning the pattern from labeled examples.
Supervised — learn from answers
You show it examples where you already know the right answer. It learns the pattern connecting inputs to outputs, like a new hire studying past cases.
Unsupervised — find hidden structure
No answers given. It looks at raw data and groups similar things together, surfacing patterns you didn't explicitly ask for.
Reinforcement — learn by doing
No labeled examples at all. It takes actions, sees the results, and gradually favors whatever actions lead to better outcomes.
Technical AI concepts, for business leaders
A plain-English map of the AI terms that show up in meetings.
A history of AI, 1950–2026
How these ideas evolved from early research to today's tools.
Try the Neural Network Playground →
See how the model that's doing the "learning" is built.
Try the Generative AI tool →
See the same idea applied to writing text, word by word.
Supervised learning learns from examples that already have the right answer attached — like a spreadsheet of past decisions and their outcomes. Unsupervised learning gets raw data with no answers at all and has to find its own groupings or patterns. Reinforcement learning learns by trial and error, taking actions and getting rewarded or penalized for the results — like learning a game by playing it.
A mix, at different stages. The core language ability comes from something closer to supervised learning at massive scale — predicting the next word from huge amounts of text. Then a reinforcement-learning-style step (learning from human feedback) fine-tunes the model's behavior to be more helpful and less harmful.
Spam filters (labeled as "spam" or "not spam"), credit approval models trained on past applications and outcomes, and demand forecasting trained on historical sales — anywhere you have past examples with known correct answers.
Customer segmentation (grouping customers by behavior without pre-set categories), anomaly detection in fraud systems, and topic discovery in large document collections — anywhere you want the data to reveal its own structure.
Recommendation systems that adjust based on what you click, pricing algorithms that adjust based on what sells, and famously, game-playing AI and robotics — anywhere an action leads to a measurable outcome that can be improved over repeated tries.
Tokenizer Playground
Type anything and watch it split into tokens live — the units LLMs actually read and get billed for.
Context Window Visualizer
See how much of a model's context window your text fills up, and what happens when you run out of room.
FROG in a Bowl Prompt Builder
Fill in Format, Role, Objective, Goal, and Context — get a copy-ready structured prompt in seconds.
Neural Network Playground
Drag two sliders and watch a real, tiny neural network turn them into a decision, live.
Generative AI Playground
Watch AI write one word at a time by predicting what's most likely to come next.
RAG Playground
Ask a question, watch notes get retrieved, then see a grounded answer versus a hallucination.
Embedding Map
Click two words and see why similar meanings sit near each other — the idea behind vector search.
Attention Visualizer
Click a word and see which others a toy transformer looks at — including the classic “it” puzzle.
Prompt Injection Lab
Watch a pasted email try to override a support agent, then flip a switch that treats it as data.
