You're already good at predicting.
Before learning how AI works, try the kind of prediction a language model calculates.
Research & verificationTransformers and next-token prediction
Publication review: Peer-reviewed conference publication that established the Transformer architecture. GPT-2 autoregressive decoding is a later development.
Independent verification: The step-by-step explanations summarize a typical autoregressive pipeline; they are not independently audited claims about any single deployed model.
Peer review of a paper does not certify this demo.
Follow one sentence.
See every transformation.
Your prompt stays with you from text to prediction. Watch the numbers change and inspect each operation.
One topic.
Three ways to understand it.
Choose a level in the header. Your explanations and some diagrams adapt as you explore.
Try your own explanation level ↗Make the invisible easy to picture.
Connect the pictures to real math.
Inspect the formulas and limitations.
What if you change the rules?
Train a small network, add neurons, adjust learning rates, and test real next-token predictions.
Open the experiments ↗ ↗A paper and a demo aren't the same thing.
Every experiment links to research and clearly states whether its publication was reviewed and its code independently audited.
Review the sources ↗ ↗CLAIMS SECOND.
Explore the science behind AI.
These plain-English guides explain how language models break up text, represent tokens with vectors, combine context through attention, and learn by updating weights. Each topic includes research references, real limitations, and an experiment you can try.