Death Plan Switch emails solo devs once a week, and after 144 hours of silence it sends their pre-written messages

Chris at a desk pressing a big round button on a mantel clock while an envelope waits sealed on the corner of the desk — Illustration: LockerGnome

The builder is Peter, who runs Tecky, a one-person web studio in Macksville, New South Wales. His clients’ sites, domains and hosting all go through him. “If I died tomorrow, nobody would know where anything was, or even that they needed to look,” he writes on the project’s page.

He already had the plan. Everything the business runs on sits in one encrypted document; his backup developer holds the password and his lawyer holds a second copy. “What was missing was the part in the middle,” he writes: something that notices he’s gone quiet and tells the right people, in the right order, “without frightening my family over a long weekend.”

So the switch counts everything from the weekly email. A reminder goes out at hour 24, to both your addresses, and a last one at hour 48. At hour 72 your next of kin gets one question, “Is everything all right?”, with two buttons: All fine or Something’s wrong. They get a reminder at hour 120. At hour 144, if nobody has said you’re well, every confirmed contact receives their message. Because the check-in goes out on the same day and hour each week, the README puts the worst case at 13 days from your last check-in to your contacts’ inboxes.

The details are where it gets thoughtful. Opening the link never counts, because email security scanners open links on their own; you have to press the button. All fine from your next of kin ends it, and nobody else hears a thing. If you turn out to be alive after the messages fire, you check in, and everyone who was contacted gets a false-alarm notice. Contacts have to opt in and confirm their role before they are ever written to, and the switch won’t arm while a message still has its “[put the pointer here]” gap in it.

It’s one PHP 8.1 app with no database: four JSON files in a data/ folder, written under a lock, with a .bak of each. A cron line runs every 15 minutes, and the installer writes that line for you. It’s MIT licensed, and pull requests are turned off: “this is a security-sensitive tool, and outside code is not merged,” the README says. Fork it if you want it different.

Peter has been running it on himself for about a week, with his next of kin and backup developer confirmed, and he’s asking anyone who trials it to tell him “what was confusing, what broke, and what you’d want it to do differently.” The repo went up October 8. Nobody outside his own install has watched it fire for real yet.

Why it matters

Solo operators are a bus factor of one. A freelancer who holds a client’s domain, DNS and hosting logins is the only key to that client’s business. When that person stops answering, the client usually finds out when a domain lapses or a certificate expires, weeks later, with nobody to call.

The scheduler is the single point of failure, and the README says so. If cron dies, the switch never fires and nothing inside it can tell you. Its answers are blunt: the weekly email doubles as a heartbeat (“Your switch is armed”), every page warns when the scheduler hasn’t run in a day, and a secret health address returns 200 or 503 for an outside uptime monitor to watch. It also wants hosting separate from the sites it protects, so the outage that takes your clients down can’t take the switch with it.

It’s honest about what it can’t defend. Anyone who can read your mailbox can check in for you and keep it quiet forever. Your SMTP password and the text of your messages sit in plain text in the data folder, which is exactly why the messages are meant to hold pointers, never logins. After a long outage it won’t dump every overdue step at once; steps go one at a time, at least 12 hours apart, and nobody is contacted before you’ve had your reminders.

Steal this

Write the pointer, keep the secret elsewhere. When you set up any “if I’m gone” message, the message says where the document is and who holds the key, never the password itself. This is like leaving a note that says “the spare key is with the neighbor at 14” instead of taping the key to the door.
Rehearse the escalation before you trust it. The repo ships tests/timeshift.php, which makes an install believe time has passed so you can walk the whole ladder in an afternoon. Do it with addresses you control, then check in and watch the false-alarm notices go out. Kind of like a fire drill: you want to find out the exit is locked on a Tuesday, not during the fire.
Watch the watcher from outside. Any scheduled job that protects you should have something off-server that notices when it stops. Point a free uptime monitor at the switch’s health address the day you install it. This is like a smoke detector with a battery chirp: the device telling you it’s dying matters as much as the alarm.
Going on a long trip? Pause it, don’t kill it. The switch pauses for up to 28 days and restarts by itself, so it can’t be left off by accident. Kind of like a vacation hold on your mail that ends on its own.

Small, self-hosted tools for one-person problems tend to stay one-person tools, and this one is built to: one install, one owner, no accounts, no hosted version. The likelier next step is the boring paperwork around it. Expect contracts for freelance web work to start asking who the backup developer is and how they’d find out they’re needed, the way leases ask for an emergency contact. Whether a 13-day worst case is fast enough for a client whose domain renews next week is the question each install answers for itself.

Search

Death Plan Switch emails solo devs once a week, and after 144 hours of silence it sends their pre-written messages

More Software