Your AI Built the App. This Hour Keeps It Alive.

Seven checks that keep an AI-built app alive - no engineering background required. Companion to "Code Is Cheap Now. Software Still Isn't." Read the full article

First, two questions.

Does your app touch other people's data or money? Would anyone notice within an hour if it broke?

Two nos: relax. You have a weekend project. Keep vibecoding - it's the cheapest way to test an idea ever invented, and nothing below applies to you yet.

Any yes: you're running production software, whether you meant to or not. The good news is that the first hour of defense doesn't require an engineer. Here are seven things to do this week. Each one exists because a real company got hurt skipping it.

1. Turn on error alerts. Right now, your app may be failing quietly - errors happening with nothing and nobody watching. Any error-tracking tool's free tier fixes this in about ten minutes. Ask your AI assistant: "Add error tracking to this app using a free tier tool, and set it to email me on new errors."

2. Set a billing alarm on every service you pay for. An artists' app once went viral - the dream - and got a $96,280 hosting bill in one week. Nothing was hacked; the pricing just did what the pricing page said. Log into every service your app uses (hosting, database, AI) and set a spending alert at about twice your normal monthly bill.

3. Restore a backup. Actually restore it. A backup you've never restored is a guess. When an AI coding agent deleted a company's production database last year, it also claimed the data was unrecoverable - it was wrong, but nobody knew until they tried. Ask your AI: "Show me, step by step, how to restore this app's database from the most recent backup into a test copy." Then do it.

4. Turn on two-factor sign-in. At minimum for you and anyone who can see user data or touch settings. Five minutes per account, and it's the single cheapest way to keep a stolen password from becoming a breach.

5. Ask your AI where the keys live. Paste this, word for word: "List every API key, password, or secret in this codebase and where it's stored. Is any of it visible to users or included in the app people download? If yes, fix it and show me what you changed." One founder skipped this; attackers found his keys within 48 hours of launch and ran up his bills until he shut the product down.

6. Write the list of every service that touches your users' data. Hosting, database, email, AI - all of it, in one document. This list is the first thing privacy law, enterprise customers, and regulators ask for. (A healthcare app we assessed had four such services and zero of the required agreements signed - while real patient data flowed through all four.)

7. Write down who answers when it breaks. A name, not a tool. One AI company let a support bot answer alone; it invented a policy that didn't exist, users cancelled in public, and a cofounder spent the weekend apologizing. If the name is yours, that's fine - write it down anyway.

One more number, for your budget: running a real product properly costs at least ~$500 a month once you have traction, and $5,000-20,000 a month by the time you have 10,000+ users - before you pay a single person. If your plan says your only cost is an AI subscription, your plan is wrong. The full article has the line-item bill.

Where this list comes from: it's the first hour of Production Bridge, the assessment we run at Kepler on AI-built apps - thirty-plus checks deep, with a keep-or-rebuild verdict, a hardening timeline, and a projection of what your app will cost to run at your scale. The difference between this page and the full assessment is the difference between locking your door and having someone walk the building. Start with the door.

Read the full article · Production Bridge · See a sample report