Braintake is a small vocabulary doing a lot of work. Here's the whole thing in one page - what each piece is, how they fit together, and what happens when you forward in your first email.
Sources flow into Intake → Drafts wait for a human → Memories update → Awareness watches the whole brain and feeds findings back into Review.
If you understand these, you understand Braintake.
A living markdown doc with one job.
The atomic unit of your brain. Each memory holds one decision, one fact, or one rule, written in plain markdown, versioned, with the people and sources that put it there. Memories aren't notes; they update over time and Braintake tracks every version with the reason it changed.
A scoped brain for a project, client or domain.
Inside one workspace you can keep multiple Braintakes: one per project, client, or area. Each is its own scope: its own memories, its own awareness, its own review queue. People can belong to more than one. Think folders, but the awareness engine respects the boundary so a contradiction in 'Client A' doesn't pollute 'Client B'.
The receipt for every memory.
Every memory traces back to where the claim came from - a Slack thread, an email, a meeting transcript, a PDF, a URL, a manual note. Sources stay attached, viewable inline, so 'because I said so' never happens. When a source is updated or removed, the memory it backs flags for review.
The doorway every new fact comes through.
Anything you forward in, paste, upload, or connect gets parsed against the brain - not just summarized. New facts in an empty area land directly (with one-click undo). Anything that touches or contradicts an accepted memory goes to the review queue with a clear before-and-after.
A proposed memory waiting for a human.
Drafts are the AI's polite way of saying 'I think this should change'. Each draft shows what's proposed, what it would update, why, and the original source. Accept and it becomes a new version; reject and it disappears. Nothing changes the canonical memory without a human nod.
Every memory's full history, with the why.
Every accepted change creates a new version. The previous one stays - title, body, author, source, and the reason it changed. Open any memory and scrub through time. Roll back if you regret it. Audit trail without thinking about audit trails.
The watcher that keeps the brain honest.
Braintake reads its own contents continuously and flags four things: stale memories no one has touched in a while, conflicts between memories that contradict each other, gaps where a topic keeps coming up but no decision exists, and open questions that were asked but never answered. Each finding is one click away from resolution.
Five minutes of daily housekeeping.
The single inbox for everything Braintake wants from you - drafts to accept, questions to answer, stale memories to confirm, conflicts to resolve. Open with your coffee, work through three or four items, close. The brain stays sharp because the queue stays short.
Query the brain in plain language.
Ask anything - Braintake answers from your memories, cites the exact ones it used, and shows confidence. If it doesn't know, it says so and offers to capture a memory once you do. Answers come from your brain, not the internet.
You send an email to your private Braintake address.
Braintake reads it against your existing memories.
One new memory + one update to 'Pricing' - both proposed.
Coffee. Click. Accept the new one, accept the update with a note.
New version of 'Pricing' is live. Old version preserved. Source attached.
A doc is a static thing you write once. A memory is a living claim: it has a confidence level, a source, a version history, and Braintake watches it for staleness and conflicts. Memories update as the world updates.
Folders are organisational. Collections (Braintakes) are scopes - the awareness engine respects the boundary, so a contradiction in Client A's brain doesn't bleed into Client B's. One workspace, many Braintakes, the same team across them.
They stay forever. Open any memory and scroll the history. Each version has the author, the source that triggered the change, and the reason - you can roll back to any of them with one click.
No. The AI proposes drafts. Drafts only become memories when a human accepts them. The only thing the AI does silently is land a brand-new memory in an empty area - and even that has one-click undo.
Set up in under five minutes. Your Braintake starts learning from message one.