AI agents — the loop as a state machine, context as a budget¶
Chapter 19 got you to a model that calls functions in a loop — ReAct, tool-calling schemas, an informal notion of a context budget — and stopped exactly where production engineering begins. A demo that chains three tool calls and prints an answer is not an agent system; it is a prompt with a for loop wrapped around it, and the gap between the two is measured in composed failure probability, a prompt cost that grows quadratically with trajectory length, and a context window that fills up whether or not anyone is watching it. …
🔒 La suite est en accès freemium — lecture complète 100 % gratuite
Tu lis ici l'aperçu libre. Le reste du chapitre (code, schémas, maths, exercices) fait partie du livre complet : crée un compte gratuit (30 secondes, aucun paiement) pour tout lire.