As LLM vendors like ChatGPT and Grok roll out memory, its useful to consider how memory becomes a moat and encourages users to stop flip flopping from one model provider to another.