Generative AI Engineering Tutorial #7 | Why chatbots forget, and how to give them memory

Generative AI Engineering Tutorial: Episode 7 of SyntaxPilot's free Generative AI Engineering full course, with real-life examples for every concept. Why chatbots forget, and how to give them memory. In this video you'll learn: • Stateless APIs: why the model forgets between calls • Conversation history: send everything, every time • Why long chats get slow and expensive • Context window and context length • The input plus output token budget • Truncation and the sliding-window history • Summarising old turns • Assistant prefill, and why newer Claude models reject it • Context rot: more context isn't automatically better Every concept starts with a real-life example, then an animation and notes, then live code in the SyntaxPilot online compiler. Practice free in the SyntaxPilot compiler (nothing to install): https://syntaxpilot-learning-2026.web... Detailed notes for this episode (code, outputs, practice): https://syntaxpilot-learning-2026.web... Chapters 00:00 Welcome 00:20 Today's agenda 00:42 The bot that forgets 01:37 Stateless APIs 02:31 Conversation history 03:03 Why long chats cost more 04:15 The context window 06:01 Real context windows 06:40 Truncation and the sliding window 07:34 Summarising old turns 09:16 Assistant prefill 10:04 Context rot 10:37 Common mistakes 11:08 Mini project: ChatSession 11:52 Quick quiz 12:24 Recap and homework 12:58 Next episode Next episode: Streaming responses Post your homework and quiz answer in the comments, and subscribe for the full basic-to-master playlist, free. #GenerativeAI #Python #SyntaxPilot