Flying High with Flutter The AI Pocket Book with Emmanuel Maggiori
Nov 26, 2025
Emmanuel Maggiori is an AI researcher and author whose work demystifies the intricacies of Large Language Models (LLMs). He explains how tokens are the building blocks of AI language processing and elaborates on context windows and embeddings that enable models to understand meaning. Emmanuel delves into the controversial training methods, including Reinforcement Learning with Human Feedback, and discusses the potential pitfalls like AI hallucinations and copyright issues. His insights make complex AI concepts accessible for developers.
AI Snips
Chapters
Books
Transcript
Episode notes
LLM Versus The App Layer
- A language model (LLM) predicts the next token and is wrapped by apps like ChatGPT to produce full responses.
- Wrappers add tricks like feeding outputs back, external searches, and function calls to extend LLM behavior.
What Tokens Really Are
- Tokens are vocabulary units: whole words, common subwords, or single letters when needed.
- Tokenization determines cost and can make non-English inputs use more tokens and cost more.
Limits Of The Context Window
- The context window is the fixed number of tokens the transformer reads at once to make predictions.
- If inputs exceed it, earlier parts may be cropped and omitted from the model's reasoning.
