Brain
Reviewing what it proposes
New pages wait in a review queue where owners and admins approve, edit or reject them — the gate that keeps a knowledge base from filling up with plausible noise.
Teaching produces proposals, not pages. The Review queue is where they become one or the other.
What the queue shows
Each teach run arrives as a batch with what it produced:
| Group | Meaning |
|---|---|
| New pages | Pages that do not exist yet |
| Already knew | Material that matched pages you already have |
| Awaiting review | Everything still undecided |
The “already knew” group matters more than it looks: it is the evidence that a second upload of the same handbook did not create a parallel set of pages.
What you can do with a proposal
Approve it — the page is created, its links are wired into the graph, and it becomes answerable immediately.
Edit, then approve. This is the important one. A proposal is a draft: retitle it, cut the padding, fix the type, split it. The moment to turn a forty-page handbook into the twelve questions it actually answers is here, not later.
Reject it. Nothing is created. Rejecting noise is not a failure of the process, it is the process.
Who reviews
Workspace owners and admins. Anyone can teach the brain; deciding what becomes shared organisational truth is a narrower right.
Turning the gate on and off
Review is a setting — Require review before saving in Brain settings.
| Setting | Use when |
|---|---|
| On | Several people teach it, or the brain is used for anything consequential |
| Off | A small team getting started, where the friction costs more than the mistakes |
With it off, teaching writes pages directly. They are still editable and deletable — the difference is whether a person looks before or after.
Reviewing well
Review in small batches, promptly. A queue of two hundred proposals gets approved wholesale, which is the same as not having a gate.
Judge findability, not prose. The question is “will this answer the question someone will ask?” — not whether the writing is elegant.
One page, one subject. A page that covers leave, expenses and notice periods will be retrieved for all three and answer none of them well.
Reject duplicates rather than merging by hand. If a proposal overlaps a good existing page, reject it and edit the original — one page beats two that disagree.