// SYSTEM NOTES
What this is, and the honest reason it exists
SmithMesh is the home network behind paraN(oid)VR — a self-hosted, isolated, self-healing camera system that watches over one house and answers to no one's cloud. The map above is its live wiring: every device, every cable, every data path, and the rules that keep the cameras walled off from everything that matters. Here's what it does, how it's built, and why it goes as far as it does.
FAQ
What am I actually looking at?
A map of the home camera network — every camera, access point, switch, and server in the house, plus the paths between them. The moving dots are packets: camera video flowing in, the wall view and single-camera streams flowing out to phones and the in-house monitor, the cellular path for when someone's away, and the TV's coax run. Toggle any flow on or off in the deck.
Is it real, or just a pretty animation?
The topology is real — it's the actual gear and wiring, derived from a single source-of-truth file. The packet motion is representative: it shows how data moves, not a literal live capture of current throughput. A real-time telemetry feed is on the roadmap.
Why are there two separate networks?
Because cameras are the least-trustworthy devices in any house, and they should never share a network with the things that matter. The cameras live in a locked zone — their own VLAN behind an Omada switch — and they can only ever talk to one thing: the server's camera-side network card. They can't reach the internet, can't reach each other, and can't reach a single device on the normal home network. The home network, in turn, has no route into the camera zone.
If the networks are separated, how does video reach my phone?
The server is the only link — and it isn't really a bridge, it's a translator. It has two network cards that are deliberately not connected to each other: no IP forwarding, no routing, no relay. Software reads the camera feeds on one side, builds a fresh composite wall view or a single-camera stream, and publishes that new output on the other side. Nothing is passed through — it's consumed and rebuilt. We call it an application-brokered render gap: not a literal air gap, but the same safety in practice.
Why not just buy a Ring or a Nest?
Off-the-shelf cameras ship your video to someone else's cloud, depend on that company staying in business and on your internet staying up, and put your footage under someone else's access policy. This keeps every frame in the house, works with the internet down, and answers to no terms of service. It also doesn't sell the inside of my home back to me as a subscription.
What happens when the internet goes out?
Almost nothing changes inside the house. The cameras, the server, the wall view, and Marci's monitor all run on the local network — none of it needs the internet to record or to display. Footage keeps writing to disk and the wall keeps showing live, connection or no connection. The only thing that needs the internet is remote viewing from a phone that's away from the house.
What if the camera VLAN itself goes down?
Then it fails safe, not dark. The cameras' first home is the isolated VLAN — the preferred, fully-walled-off mode. But if that isolation layer drops, the system falls back to LAN local-only as a backup: the cameras keep recording and keep showing inside the house on the plain local network instead. Two things never break, no matter what — you can always see and record inside, and the cameras stay local-only, never exposed to the internet. The only thing that relaxes in fallback is the wall between the cameras and the rest of the home network, and it's restored the moment the VLAN comes back. The priority is deliberate: never go blind. Keep eyes on the house first, restore full isolation second.
Can the cameras be hacked?
Nothing is unhackable, and pretending otherwise is how people get burned. So the design assumes a camera could be compromised and makes that nearly worthless: a hijacked camera is trapped in an isolated zone with no internet, no path to any personal device, and no way to talk to the other cameras. The blast radius is one camera shouting into a wall.
Why does the TV / set-top box show up on a security map?
Because it shares the house's physical coax, bridged onto the network with a MoCA adapter — and an honest map shows everything that's actually on the wire. It also documents a running joke: a 65-inch 4K TV being fed about 2K by a set-top box that can barely keep up, sitting next to a paid 4K streaming subscription nobody uses.
What's with the Wi-Fi names?
Two perimeter access points, named fuckouttahere and Whouthinkurhoe.
They sit at opposite edges of the property on opposite ends of the 2.4 GHz band (channels 1
and 11) so they never step on each other or on the interior mesh, which settles into channel 6.
And if you have to name a thing, name it something that makes you smile — an SSID that reads like
a warning is a small, free deterrent.
Does it really fix itself?
Yes. The recording and streaming pieces are watched by a process that detects stalls and restarts the failed component on a staged retry, with grace periods so a brief hiccup doesn't trigger a needless restart. The goal is a system that recovers from its own faults without anyone touching it — because the entire point is that it keeps working when no one's home.
Why build it instead of buying it?
Because building it means I understand every inch of it, I can fix it at 3 a.m. without a support ticket, it does exactly what I want and nothing I don't, and no part of it depends on a company deciding to keep a product alive. And because the building is the point — keep reading.
Why this is overkill
Plainly: it is. Nine cameras, a dedicated server, an isolated camera VLAN, dual network cards that refuse to talk to each other, a self-healing watchdog — by any rational threat model this is far more than a suburban house needs. Murrieta is the sixth-safest city in the country. No burglar risk on earth justifies this much engineering. If you measured it against the actual odds of something happening, you'd tear most of it out.
So don't measure it against the odds. Measure it against what it's protecting.
Why it's necessary — for Marci
Here's the honest version.
I didn't build this for a threat. I built it for my mom.
Marci lives with me. She's recovering, and I'm gone a lot — I drive for a living, a thousand miles some weeks, so there are long stretches where I'm three counties away and she's in the house alone. I can't be in two places. So I built something that can be where I can't.
That's what the overkill actually is. It isn't paranoia — it's margin. When the thing you're protecting is irreplaceable, you don't build for the average day. You build for the day everything else has already failed, and you make sure this still holds. You add the extra camera. You wall the cameras off so they can never become a way into her life, and her life can never become a way into the cameras — watched over without being watched, safe without being exposed. You make the whole thing heal itself, because a system that needs me to babysit it is a system that's dark the moment I'm not home, and not-home is the exact problem I'm solving.
I learned what it means to keep watch over someone while taking care of my dad. The lesson was that presence is the whole thing — that the most important kind of help is just competent, uncomplaining, always-there attention. I can't give her that in person every hour of every day. So I gave it to her in the only other form I had: something that keeps watch when I can't, never gets tired, never looks away, and quietly fixes itself so it never goes dark.
The ridiculous network names, the over-the-top ops dashboard, the whole production of it — that's just the wrapping. Underneath it is the plainest thing there is. I want my mother safe, and I want to be able to glance at my phone from a parking lot ninety miles away and see that the gate's shut, the back patio's clear, and she's okay.
That isn't overkill. For her, it's exactly the right amount.