Thinking in Tokens
A conceptual mental model for modern AI systems — tokens, embeddings, attention, transformers, hallucination, retrieval, and agents — written with no math and no ML background required.
Thinking in Tokens is a conceptual introduction to how modern AI actually works — tokens, embeddings, attention, transformers, training, hallucination, retrieval, tool use, agents, reasoning models — written for curious, intelligent readers with no math, no programming, and no prior ML background.
Most explanations of AI are either marketing (“it’s magic”) or academic (equations first, intuition never). This builds one cumulative mental model, incrementally, instead of presenting AI as a list of disconnected techniques and product names — one that keeps working as the technology changes.