Interactive animations of modern language models.
Modern AI systems consist of billions of interactions that can be described mathematically — but are hard to grasp intuitively.
These interactive animations make the central mechanisms of modern language models visible step by step — controllable in time, technically precise, and without unnecessary simplification.
As if under a microscope, a view opens onto processes that normally remain hidden.
How does AI recognise patterns?
Neural networks process signals, weight information and detect relationships.
Open animation →
How does a neural network learn?
The animation shows how errors are propagated back and weights are adjusted.
Open animation →
Why does AI need GPUs?
Why modern language models would barely work without massive parallelisation.
Open animation →
How does AI understand context?
Attention reveals how language models calculate relationships between words.
Open animation →
How does language emerge?
Token by token, new text emerges from probability distributions.
Open animation →Between mathematics and imagination
The individual mechanisms behind modern AI can still mostly be explained.
But their interplay produces emergent dynamics that even experts can no longer fully grasp intuitively.
These animations decompose the processes into observable steps — not to fully "decode" AI, but to make the boundary of our understanding visible.
Who are the animations for?
For anyone who wants not only to use AI, but to understand it.
The content is aimed at learners, teachers, developers, creatives, decision-makers and curious minds who want to look beneath the surface of modern language models — without having to study computer science.
The animations are free to use in classrooms, talks or your own training sessions. For questions or suggestions for further topics, I welcome your feedback.