Every piece of tech in this house is capable of watching, listening, or logging something. That's not a design flaw — it's mostly the point. But it means every automation I build carries a quiet question I try never to skip: am I building this to look after my family, or to keep tabs on them? Those two things can look identical from the outside and be completely different in practice.
With a busy household — kids coming and going, someone always halfway out the door — it's tempting to want eyes on everything. But "I could see it if I wanted to" and "I'm going to watch it" are not the same commitment, and the gap between them is where trust actually lives.
The camera test
Here's the rule I've settled on: a camera earns its place if it's answering "is everyone safe" — the driveway when a car pulls in, the pool fence, the front door at night. It does not earn its place if it's really answering "what is my teenager doing right now." The first is care. The second is surveillance wearing care's clothes, and kids can tell the difference even when they can't articulate it.
Practically, that means: cameras point outward and at shared spaces, not into anyone's room or at anyone's screen. Footage exists to answer a specific safety question, not to be idly scrolled through. And everyone in the house knows what's recorded and why — nothing quietly watching that hasn't been talked about at the dinner table.
Automation is easier to trust than a person with a laptop
There's something useful about letting the automation itself be the boundary, rather than my own restraint in the moment. If a rule only ever fires "notify me if the pool gate opens after 9pm," it structurally can't be used to check up on anything else — I don't have to trust myself not to go looking, because there's nothing else to find. Building narrow, single-purpose automations instead of broad "just in case" monitoring is a benevolence decision as much as a technical one.
The same question, aimed at the AI layer
This matters just as much now that there's a memory system quietly learning how this house works. It would be easy to let it remember everything about everyone, all the time, "just in case it's useful." I'd rather it remember the house's patterns — energy, schedules, what worked — than turn into a running log of what everyone in the family did today. Useful and watchful are not the same thing, and I want to keep choosing useful.
None of this is a solved problem, and I expect I'll get the line wrong in both directions at different points. But I'd rather keep asking the question out loud than assume good intentions are enough on their own.