00 / Private memory layer / v0.1

One memory.Every model.

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.

CLAUDEREAD · WRITECHATGPTREAD · WRITEGEMINIREAD · WRITECURSORREAD · WRITEPERPLEXITYREADOLLAMALOCAL · OFFLINEFIG. 01 — SHARED MEMORY GRAPH6 SOURCES → 1 MIND
Models linked
06
Shared memory
01
Storage
Local
Times re-explained
00.

01 · How it works

Write once.
Remember everywhere.

Every model I use plugs into the same store. What one learns, the others already know.

Step 1 of 3

Capture.

Conversations leave a trace. Decisions, preferences and context are distilled into short, plain memories as I work.

Step 2 of 3

Connect.

Each model plugs in through one protocol. Claude, ChatGPT, Gemini, Cursor — they all speak to the same store.

Step 3 of 3

Recall.

Ask anywhere, answer with everything. The right memories surface in whichever model I happen to be talking to.

02 · The log

Everything it knows, in order.

Each memory is one line of text: when, from which model, what kind, what was decided. Readable, editable, deletable.

Sample entriesWriting
TimeSourceKindMemory
19:15CLAUDE● decisionLanding page: minimal, line art, no purple.
17:30CHATGPT○ preferenceShort answers. Code over prose.
14:44CLAUDE● decisionEvery model can read. Only I can delete.
11:20CURSOR○ contextThe brain repo is the source of truth for its own spec.
09:02OLLAMA○ contextOffline notes sync when the connection is back.

03 · Principles

Stop re-explaining yourself.

Private

Only the models I connect can read it. Nothing is shared by default.

Portable

Plain text in, plain text out. No model owns my context.

Permanent

Context outlives any single chat, tab or subscription.