AI memory should belong to the user, not the AI app

The more AI tools I used, the stranger their memory systems started to feel.

Claude knew one part of my life. ChatGPT knew another. Cursor understood a project until I switched tools. Every new assistant meant rebuilding context from scratch.

I do not think the answer is choosing one AI and putting everything inside it.

I think memory should be its own layer.

That is the idea behind Second Brain v2: one user-owned memory that Claude, ChatGPT, Cursor, Codex, and other tools can share.

The AI can change. The interface can change. Your context stays with you.

In v2, that memory is no longer just a collection of notes. Related memories organize themselves into a knowledge graph, so recall can follow the relationships between projects, decisions, people, ideas, and later updates.

The difficult part has not been storing more information. It has been making recall useful and predictable.

A direct match should still matter more than something several connections away. A trusted memory should not be silently overwritten by a conflicting one. The graph should help surface context without turning into an unexplainable pile of associations.

That is the direction I have been working toward.

Second Brain v2 launches tomorrow.

Before it does, I am curious: would you rather your AI’s memory live inside your favorite AI product, or exist independently so you could take it between tools?