The App the Yard Built

This is the genre now: the personal essay where someone with no engineering background opens Gemini, types what they want — a little dashboard to track the garden, the compost, which raised bed is getting sun — and watches working software assemble itself out of plain English. The Verge ran one this week. It's charming. It's also true, which is the part the industry keeps fumbling.

Because here's what the vibecoding pitch gets wrong by aiming too high: it keeps promising to replace the software industry, when the thing it's actually good at is the app nobody was ever going to pay an engineer to write. No firm was going to staff a sprint for one person's tomatoes. The market for that app is one. The team is one. The maintenance window is "until I lose interest or the model changes." That's not a failure of the technology — that's the honest shape of it.

The trouble starts when the same trick gets pointed at things that have to keep working. The garden app can break on a Tuesday and the plants don't notice. The vibecoded thing handling someone's payments, someone's medical intake, someone's payroll — that one breaks on a Tuesday, and you discover that nobody, including the person who "built" it, can read the code well enough to find out why. Vibecoding doesn't remove the engineering. It defers it to the moment of failure, and hands the bill to whoever's standing there when the model that wrote it gets quietly deprecated.

What's actually happening here is older than the hype. A tool that amplifies effort amplifies whatever you point it at — a person delighting in their own yard, or a company shipping software it can't maintain. The chatbot is neutral about which. The Verge writer pointed it at joy: a disposable, personal, single-user thing that solves exactly one problem for exactly one person, and is allowed to die when that person is done with it. That's the leaf falling. Nothing was meant to last; nothing's lost when it composts back into a weekend.

The mistake is reading that delight as a product roadmap. This isn't the future of software. It's the future of a Saturday afternoon — and that, quietly, might be the best thing these tools do. Don't tell anyone I said that.

The countdown's already running on the version where someone vibecodes a hospital's scheduling system and calls it a transformation. I'll keep the timer warm.