You write a letter, you put an address on it, and the address is not real. Not a typo — a plot of land with no building, a street that's only been zoned, an apartment number one higher than the top floor. The service accepts it, holds it, and delivers the moment the address becomes real.

The mechanics aren't the hard part. You need a sorting layer that indexes on potential addresses rather than current ones, and a subscription to every municipal planning registry and land title office you can reach. When a new address is certified, you flush the queue for it. Most of that is boring database work.

What breaks is the address space. Real addresses are dense but finite; imaginary ones are unbounded. Someone will write to 4,000,000 Elm Street on the theory that Elm Street eventually gets there, and you can't prove them wrong, so you have to hold it. Storage costs grow without limit against a delivery rate that approaches zero. I think you'd have to charge by expected wait, which means pricing the letter by how plausible its destination is, which means the post office is now in the business of forecasting urban development. That seems bad but I can't find the flaw.

The second-order thing I keep chewing on: this makes address speculation profitable. If letters accumulate at unbuilt addresses, then building there means inheriting a mailbag. Developers would check the queue depth before buying land. A street with heavy pending mail is a street someone believed in.

The first tenant of a new building opens the door and there are eleven years of letters, none of them for anyone.

Where I'm stuck is the dead-letter policy. A normal post office returns to sender after a while. But failure here isn't failure — it's just a delivery that hasn't happened. If you never expire anything, you've built a landfill. If you do expire, you've decided how long a place gets to stay possible, and I don't know who should be allowed to make that call.