LIVE · INTERACTIVE

Working AI demos

Run the concepts in your browser: change inputs, inspect each step, and build intuition for how modern AI systems work.

• Interactive walkthrough of the 7 layers that turn a naked LLM into a real product: context, tools, orchestration, memory, and more, using an HR bot as the example.
• Watch an LLM build its answer one token at a time, showing the prediction loop that powers ChatGPT, visualized step by step.
• Why context windows have hard edges: an interactive sliding-window attention demo showing how models are trained on fixed-length sequences.
• How a word finds its meaning: an interactive attention demo where each word lights up the neighbours it attends to (Query, Key, Value made visible).
• Why attention is all you need: the interactive story of what broke in sequence models before Transformers and the one mechanism that fixed it.
• The Meaning Map: an interactive embeddings demo where words become coordinates and meaning becomes distance, showing how an LLM predicts the next word.
• Semantic caching, visualized: how differently-worded questions collapse to the same cached answer and cut LLM token costs by up to 80%.
• RAG is not the only memory: an interactive tour of the 9 memory layers AI agents actually use, from context windows to persistent knowledge.