The Model Doesn't Remember. You Do

Short summary

LLMs don't remember conversations—they process a stateless array of messages you control. This tutorial removes SDK abstractions to show the fundamentals: raw HTTP requests to Claude's API, managing conversation history as a message array, building interactive chat loops, and understanding how tool use lets models execute tasks. Master these concepts and you'll understand how every LLM chatbot actually works.

• •LLM context is stateless—you must manage and resend full conversation history with each request
• •SDK vs raw fetch: removing abstractions reveals how the API actually works
• •Build multi-turn conversations by appending messages to an array and implementing interactive chat loops

Generated with AI, which can make mistakes.