Wait, normal people dont immediately think "I could have live ants shipped to anyone/pick up anyone's car under this system"?

Address verification has some well-established methods, like sending a OTP through the mail.

We don't answer calls anymore, and a "fake" pizza delivery, doesn't that mean the person get's a free pizza?

Here, you're disrupting someone's most vital health function for a low fee.

No joke, I won the World Sleep Championships a few years back, and received two deliveroo knocks on my door that night. I was modestly suspicious that it was intentional interference to throw me off (of course, there was no money on the line, and I am pretty sure other competitors didn't know where I lived).

https://www.affectablesleep.com/blog/neurohacking-the-world-...

You could add some kind of recourse mechanism and make customers post bond. Like if the wrong person is woken up they can visit a URL that causes the originator to be fined / lose a deposit.

This is definitely a problem to solve. I could even see harassment being the most likely initial use case, until you manage to reach people with the problem seeking a solution. (People are more likely to want to eat pizza themselves than to harass with it. But people who just heard of this door-knocking service, without seeking it out, are more likely to want to harass someone than to want to be woken up themselves.)

A lot of tech businesses try to ToS away liability, but you can't do that in this case, since the harmed party isn't the customer/user. (You can try to ToS away the liability of your door-knocker flaking on you, or the customer thinking they did, and missing an important meeting. But not the harassment of a non-customer/user.)

I don't think zero-knowledge proofs of residence are ready.

If you could find a way to do it in a smooth-UX way, such as by signin-with-Google (or confirmed email) and match that up with physical address using a creepy data-broker service, that might work well. But I'd guess would be a big percentage of your engineering effort, and you'd have ongoing costs, and possibly some upfront commitment to the broker to bother with you at a viable cost rate.

Other ideas that come to mind seem like they'd have significant numbers of rejected legitimate, and accepted illegitimate.

Random idea: One of the times people most want wakeup help is when they're traveling (with disrupted schedules, unfamiliar settings, risk of phone alarm accidentally in DND/mute or out of battery, etc.). Hotels have it covered. Maybe you could integrate with AirBnb, in a way that lets you sufficiently authenticate that the person at the address at that time wants to be woken then. And you can give AirBnb a big cut, for the integration and for advertising your service. Or maybe AirBnb wants to build and own the UI and billing, and you're only a middleperson who supplies and pays the contractor door-knockers (and provides a brand, and lets AirBnb keep a bit arms-length on that and the contractors). (Or "hosts" could provide an unusually good alarm clock on the nightstand. Or there could be an unusually good alarm clock that the people who want it can buy.)

Can’t be that much crazier than delivering food right?

Unless I’ve missed something.

How do you verify that a doorway in a photo belongs to a given address?

Could be easy to spoof with AI image gen and a telephoto of the front door of Shooty McTriggerhappy, the neighbourhood crank whom Badfaith Faux-Customer would like to see knocked up.