Step 1 of 3
Capture.
Conversations leave a trace. Decisions, preferences and context are distilled into short, plain memories as I work.
00 / Private memory layer / v0.1
A shared memory that sits under every AI I use. Context, decisions and preferences are written once — and every model reads from the same mind.
01 · How it works
Every model I use plugs into the same store. What one learns, the others already know.
Step 1 of 3
Conversations leave a trace. Decisions, preferences and context are distilled into short, plain memories as I work.
Step 2 of 3
Each model plugs in through one protocol. Claude, ChatGPT, Gemini, Cursor — they all speak to the same store.
Step 3 of 3
Ask anywhere, answer with everything. The right memories surface in whichever model I happen to be talking to.
02 · The log
Each memory is one line of text: when, from which model, what kind, what was decided. Readable, editable, deletable.
| Time | Source | Kind | Memory |
|---|---|---|---|
| 19:15 | CLAUDE | ● decision | Landing page: minimal, line art, no purple. |
| 17:30 | CHATGPT | ○ preference | Short answers. Code over prose. |
| 14:44 | CLAUDE | ● decision | Every model can read. Only I can delete. |
| 11:20 | CURSOR | ○ context | The brain repo is the source of truth for its own spec. |
| 09:02 | OLLAMA | ○ context | Offline notes sync when the connection is back. |
03 · Principles
Only the models I connect can read it. Nothing is shared by default.
Plain text in, plain text out. No model owns my context.
Context outlives any single chat, tab or subscription.