diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
index 97dc880..868e4c6 100644
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -5,10 +5,26 @@ openblox runs untrusted, machine-generated code in disposable Linux sandboxes.
 It is a **Go library first** and a daemon second. There is no control plane, no
 database, and no scheduler. A sandbox is a container; the container is the state.
 
-## Non-goals
+openblox is deliberately the layer *below* a sandbox platform. If you need
+scheduling, tenancy or an API of your own, build them on top — the absence of a
+control plane here is what makes that possible, rather than something you have
+to work around.
+
+## The rule
+
+One line decides whether something belongs in openblox:
+
+> **How a sandbox is isolated is openblox's problem.
+> Which sandbox runs where is yours.**
 
-Stating these up front, because every one of them is a place a sandbox project
-grows into a platform:
+Isolation is the runtime, the egress policy, the filesystem, the user, the
+capabilities, the resource caps and the lifetime bounds. Placement is
+scheduling, queueing, tenancy, metering, fleet membership and snapshots.
+
+Every non-goal below is an application of that rule, not a separate opinion —
+which is why the list has not grown and is not expected to.
+
+## Non-goals
 
 - **No multi-tenant SaaS.** No organizations, billing, usage metering, or audit log.
 - **No scheduler.** One host, one daemon. Not a fleet.
@@ -17,8 +33,10 @@ grows into a platform:
 - **No sub-second cold starts.** Correctness and containment over latency.
 - **No database.** Container labels are the registry.
 
-If you need those, you want a hosted platform. That is a legitimate need and
-openblox is the wrong tool for it.
+If you need those, build them on top of openblox or use a hosted platform. Both
+are legitimate; neither belongs in here. A feature request that crosses the rule
+is declined by pointing at the rule, which is the only way a project this small
+stays this small.
 
 ## Shape
 
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index f3d9ac5..8032eb8 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -1,5 +1,22 @@
 # Contributing
 
+## Scope
+
+Read this before proposing a feature. One rule decides what belongs in openblox:
+
+> **How a sandbox is isolated is openblox's problem.
+> Which sandbox runs where is yours.**
+
+Isolation is in scope: runtimes, egress policy, filesystem, users, capabilities,
+resource caps, lifetime bounds. Placement is not: scheduling, multi-node, tenancy,
+metering, snapshot/resume, an API server, a database.
+
+openblox is the layer below a sandbox platform. Anything on the placement side is
+something you can build on top, and keeping it out is what keeps this small enough
+to audit. A PR that crosses the rule will be declined on that basis, however good
+it is — so please open an issue before writing one. [ARCHITECTURE.md](ARCHITECTURE.md)
+has the reasoning.
+
 ## Setup
 
 ```bash
diff --git a/README.md b/README.md
index aa6d71a..dfa05f3 100644
--- a/README.md
+++ b/README.md
@@ -59,6 +59,30 @@ openblox takes the third option: a substrate small enough to read in an
 afternoon, that you run yourself, with isolation supplied by gVisor rather than
 by hope.
 
+## Where this sits
+
+openblox is the layer *below* a sandbox platform, not a smaller one.
+
+```
+  your scheduler, your tenancy, your API    ← yours to build, if you ever need it
+  ──────────────────────────────────────
+  openblox                                  ← isolation, done correctly
+  ──────────────────────────────────────
+  Docker + gVisor                           ← the boundary itself
+```
+
+One rule decides what belongs here:
+
+> **How a sandbox is isolated is openblox's problem.
+> Which sandbox runs where is yours.**
+
+Egress, capabilities, filesystem, resource caps, lifetime, runtime: openblox's.
+Placement, queueing, tenancy, metering, snapshots: not openblox's, and not
+planned. Build those on top when something actually asks for them. That is what
+a lower layer is for, and it is why there is no control plane to adopt first.
+
+The comparison is `libvirt`, not OpenStack.
+
 ## Secure by default
 
 The zero value of every option is the safe one. A sandbox created with no
@@ -81,6 +105,16 @@ If the host cannot provide the requested isolation, `Create` fails with
 sandbox that is quietly less isolated than you asked for is worse than no
 sandbox, because you keep trusting it.
 
+**Two levels of the same guarantee.** In the library, *your* code chooses: the
+defaults are safe, and every relaxation is explicit and greppable at the call
+site. Through [`openbloxd`](https://openblox.sh/security/#deploying-the-policy-broker-openbloxd)
+the choice stops being the caller's at all — profiles live in the daemon's
+config file and no request can reach them. A caller names a profile. It cannot
+name an image, a runtime, a user, an egress policy, or a resource cap.
+
+That is the difference between weakening being **visible** and weakening being
+**unreachable**, and it is the whole reason the daemon exists.
+
 ## What you get
 
 | | |
@@ -93,13 +127,18 @@ sandbox, because you keep trusting it.
 
 ## What it is not
 
+The first three follow from the rule above — they are placement, not isolation.
+The fourth is a trade made on purpose.
+
 - **Not multi-tenant.** No organizations, auth, billing, or metering. Tenancy is
   the caller's concern.
 - **Not a fleet.** One host, one daemon.
 - **No snapshot, fork, or pause/resume.** Stop and re-create from a baked image.
 - **Not the fastest.** Correctness and containment over cold-start latency.
 
-Those are deliberate. See [ARCHITECTURE.md](ARCHITECTURE.md) for the reasoning.
+None of these are gaps waiting to be filled. They are the boundary that keeps
+openblox small enough to be worth reading, and requests to cross it get declined
+on that basis. See [ARCHITECTURE.md](ARCHITECTURE.md) for the reasoning.
 
 ## Install
 
diff --git a/docs/index.md b/docs/index.md
index 664bb3b..c39cd86 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -39,6 +39,29 @@ kernel with the host.
 openblox takes the third option: a substrate small enough to read in an afternoon, that
 you run yourself, with isolation supplied by gVisor rather than by hope.
 
+## Where this sits
+
+openblox is the layer *below* a sandbox platform, not a smaller one.
+
+```
+  your scheduler, your tenancy, your API    ← yours to build, if you ever need it
+  ──────────────────────────────────────
+  openblox                                  ← isolation, done correctly
+  ──────────────────────────────────────
+  Docker + gVisor                           ← the boundary itself
+```
+
+!!! quote "The rule"
+    **How a sandbox is isolated is openblox's problem.
+    Which sandbox runs where is yours.**
+
+Egress, capabilities, filesystem, resource caps, lifetime, runtime: openblox's.
+Placement, queueing, tenancy, metering, snapshots: not openblox's, and not planned.
+Build those on top when something actually asks for them. That is what a lower layer
+is for, and it is why there is no control plane to adopt first.
+
+The comparison is `libvirt`, not OpenStack.
+
 ## What you get
 
 <div class="grid cards" markdown>
@@ -73,6 +96,9 @@ you run yourself, with isolation supplied by gVisor rather than by hope.
 - **Not a substitute for a threat model.** Read the [security model](security.md) and
   decide whether its guarantees match your workload.
 
+The first two are the rule above, applied: they are placement, not isolation. They
+are not gaps waiting to be filled.
+
 ## Next
 
 - [Quick start](getting-started.md) — install, prerequisites, a working example