Next LLM · Chatbot
Chatbot Systems
Not written yet — this is the placeholder for the conversational-interface layer between raw LLM calls and full autonomous agents.
placeholder
1
Overview
What belongs in this track once it's written.
◆ Planned topics
This track sits between LLM internals and full agents: it's specifically about the conversational-interface layer — turn-taking, session state, and UX around a chat-style LLM product, not autonomous tool-using behavior.
- Multi-turn conversation state and how it's threaded into each request
- System prompt / persona design for a chat assistant
- Streaming responses and perceived-latency UX
- Session memory vs long-term user memory
- Conversation summarization / context compression strategies
- Moderation and safety filtering on both user input and model output
- Handling interruptions, edits, and regenerations mid-conversation
- What would you do if the chatbot suddenly started hallucinating in production?
Fill this page in question-by-question the same way the other tracks are written, then this callout can go.