MENSARA.AI · A GOVERNED MIND, IN YOUR ACCOUNT

MENSARA.AI · THE GARDENING · MACHINE SURFACE · TEXT/MARKDOWN

# The gardening

> When something in a Mind looks stale, a change gets drafted, never applied directly. A deterministic policy check decides whether it can proceed; a model-based check only narrows and flags. Today every proposed change, everywhere, waits for the owner's explicit approval before it takes effect. Automatic merging for trusted, low-stakes categories is a real, built feature, switched off in every deployment right now. Content governing how a Mind decides and what it may do always requires human review, with no fast path at any trust level.

Canonical: https://mensara.ai/how-it-works/the-gardening

## What going stale looks like

A pricing rule changes. A process gets shorter. A person leaves and takes a piece of "how we do things" with them. None of that rings a bell on its own. What does ring a bell is something in your Mind no longer matching how the business actually runs, caught the next time it gets read, referenced, or asked about. That mismatch is the signal. It is quiet, and it is exactly the kind of thing that gets missed without a mechanism looking for it.

## Proposed, not applied

When your Mind finds something that looks stale, it drafts a change. It does not make the change. The draft sits in front of two checks before it can go anywhere near the score. The first is a rule, plainly written and not a model. It decides. The second is a more careful read, done by a model, whose job is to narrow the field and flag anything the rule should know about. It does not get the final word. The rule does.

## Nothing moves without your yes

Every proposed change, in every part of your Mind, waits for you. Today, that is not a soft default or an option you have to find and turn on. It is the only path there is. No category is exempt, no track record shortcuts it, no change goes live on its own. You read it, you decide, and only then does it take effect. That is not a limitation we are working around. It is the point.

## What earns speed, and what never does

The checks that review each change are also what, over time, build a record. That record is designed to matter. Certain low-stakes kinds of content could eventually earn the right to move faster, once they have shown, proposal after proposal, that they get judged the same way a careful owner would judge them. That path is built into the design. It is not switched on anywhere yet. And some things are built to never take the fast path at all. Whatever governs how your Mind decides, and what it is permitted to do, stays in front of a person, permanently, regardless of what the record shows.

**Status:** running since 2026-07-10 · the review pipeline is live · automatic merging is built but switched off everywhere, every change waits for your yes

## Prompt

```prompt
Check whether your Mind currently has any proposed changes waiting for your review. For each one, show me what triggered the proposal, what the policy check decided, and what the model-based review flagged. Confirm nothing has taken effect without my approval yet.
```

## Related

- [The gateway](https://mensara.ai/how-it-works/the-gateway)
- [Receipts](https://mensara.ai/how-it-works/receipts)
- [Held at the gate](https://mensara.ai/how-it-works/governance)