shogol

Series · AI

From Tokens to Agents

How modern AI actually works — from the raw text a model reads, to the engines that serve it at scale, to the agents that act on your behalf. Intuition-first, with diagrams.

  • 15 parts
  • 88 min end to end
  • 9 sections
  • Latest Jun 2, 2026

Contents

  1. 01The Big PictureWhat actually changed, and a map of the whole stack.Foundations · 5 min
  2. 02What Is a Large Language Model?Why “predict the next token” turned out to be so powerful.Foundations · 6 min
  3. 03Tokens & TokenizationHow text becomes the numbers a model can chew on.Foundations · 5 min
  4. 04Inside the TransformerAttention, embeddings, and what's really happening in there.Foundations · 5 min
  5. 05How Models Are TrainedPretraining, fine-tuning, and the human feedback loop.Training · 6 min
  6. 06Inference: How a Model Generates TextAutoregression, sampling, and temperature.Inference · 6 min
  7. 07The Inference EngineKV cache, batching, and the economics of serving.Inference · 7 min
  8. 08Making Models Faster & CheaperQuantization, distillation, speculative decoding.Inference · 6 min
  9. 09Prompting & ContextThe prompt as a program you write in English.Prompting · 6 min
  10. 10The HarnessThe loop and scaffolding that turn a model into a product.Harness · 5 min
  11. 11Tool Use & Function CallingGiving the model hands.Tools · 6 min
  12. 12What Is an Agent?The agentic loop, planning, and memory.Agents · 7 min
  13. 13Multi-Agent SystemsOrchestration and dividing the work.Agents · 5 min
  14. 14RAG & MemoryGrounding models in real, current knowledge.Memory · 6 min
  15. 15Evaluation, Safety & What's NextHow we measure quality, manage risk, and where this is heading.Future · 7 min