We wipe your server. Here's why.

"We wipe your server."

It's the sentence in the onboarding flow that makes people stop and read it twice. You're handing us a machine and the first thing we do is erase it — disk and all. That sounds like the dangerous part. It's the opposite; the wipe is the whole point.

You point us at a fresh VPS. Usually Ubuntu — that's what almost every provider hands you by default. We take it over and reformat the entire disk, then set it up as a ZFS volume. On top of that we install NixOS.

That's it. A clean box in, a known system out. We don't adopt what was there, patch around it, or work with whatever the provider pre-installed. We start from an empty disk we laid out ourselves, because everything we do afterward depends on that.

The one thing we can't do without is a baseline we actually understand — and that's NixOS. We're offering to maintain the server for you; that only works if we know, exactly, what's on it. Adopt an existing box and we inherit its whole history — every package someone installed, every config file someone edited at 2am, every leftover we can't see and can't account for. We'd be maintaining "an Ubuntu VPS somebody did things to," not a system we can reason about. So we don't adopt it. We install NixOS from scratch, where the entire machine is described in code and every change is a tracked generation — roll back a bad change, know precisely what's running and why. That's not negotiable, and there's no way to get there without wiping what was there first.

ZFS is the part we'd rather not skip. We format the disk as a ZFS volume so we get snapshots in place and backups that don't get in the way — restore a server to a known state in seconds, without the filesystem fighting us. But ZFS wants the disk laid out its way from the start; you can't sprinkle it onto a running install after the fact. On a box with less than 2GB of RAM we drop it and fall back to ext4 — and that machine still gets wiped and rebuilt on NixOS anyway. So even the stripped-down path starts from an empty disk; the wipe was never the ZFS tax.

It comes down to something pretty basic: we can't stand behind a machine we didn't build. Inherit someone's box and we're guessing about it forever; build it ourselves and we actually know.

Is this dangerous? No — because we expect a clean box. When you enroll a fresh VPS, there's nothing on it to lose; the wipe erases a machine that's effectively empty already.

The care is entirely on the front end. If there's data on that server, backing it up first is on you — we can't see it, we can't detect it, we don't go looking. We wipe what we're given. So if you've got something on there that has to survive, or the whole idea just makes you nervous, don't point us at it. Talk to us instead; we'll help you spin up a fresh VPS and move over cleanly.

The remnants are the actual risk — not the wipe. A leftover cron job, a package nobody remembers installing, a config that drifts a little further from what anyone intended every month. That's the server that fails at the worst possible time, and it's exactly what we're refusing to inherit.

A clean disk isn't us being reckless with your machine. It's the only honest starting point for taking it off your hands for good — no history we can't account for, no state we didn't build, nothing left to surprise either of us at 3am.