Read just enough
The essential idea comes first. Technical precision appears only when it becomes useful.
Understand by experimenting
AI, without the smoke and mirrors.
A free, open-source learning platform for understanding generative AI through plain explanations, local experiments, and clearly stated limits.
6 lessons · 68 min6 labsNo account100% local
Next tokenSimulation
The cat drinks
milk62 %
water24 %
tea9 %
A different way to learn
Less magic. More grip.
Each lesson starts with a concrete question, isolates one mechanism, and lets you test it. Simplifications are useful when their limits are visible.
Change something
Adjust a budget, source, or tool and observe what the mechanism actually changes.
Check the mental model
Short takeaways and optional technical notes help you confirm what changed and what did not.
01How does an LLM generate a response?7 min02What is a token?11 min03Why can two answers be different?12 min04Why can an LLM hallucinate?13 min05What is the context window for?12 min06Model, chatbot, agent: what is the difference?13 min
Path 01 · complete
Six questions. One clear thread.
Start with token-by-token generation, then move through hallucinations, context, and agents. Every step includes a practical experiment and concise takeaways.
Local by defaultProgress stays on this device.
Visible sourcesPrimary references and review dates.
Declared limitsA simulation never pretends to be a model.
A reasonable promise
Understand enough to ask better questions.
You do not need to become a machine-learning engineer. The goal is to build useful mental models, spot shortcuts, and stay critical of both answers and promises.