Ask HN: How do you keep AI assistants consistent with your personal preferences?

I'm the founder of CORE (https://github.com/redplanethq/core) - an external memory system for LLMs that acts like a "memory OS." Instead of AI starting fresh every conversation, CORE lets you create persistent knowledge that any LLM can query and reference.

We've built "spaces" as scoped memory containers - think isolated knowledge environments for different contexts. You might have a 'coding-standards' space with your preferred patterns, a 'project-context' space with current work details, or 'personal-workflows' space with how you like things done. Any AI agent can pull from the relevant spaces to give consistent, contextual responses.

But I'm trying to nail the UX where both user and agent are crystal clear about what preferences are active. How do you make it obvious when AI is referencing your stored knowledge vs. responding generically? Should users explicitly choose which spaces are active, or should AI auto-select based on context?

What personal context do you find yourself repeatedly explaining to AI assistants? And how would you want to manage and be notified about which of your knowledge spaces the AI is actually using? Looking for insights from people who deal with AI consistency challenges daily.