Hardening off is a week, not a button

The button that deserved to die

We shipped hardening off as a task. One line, one tap, done.

A gardener using the app pressed it and told us what happened, in his own words: "I clicked that button, not sure what happened but it dissapeared. beginners don't know what harden off means, and likely harden off should be a series of tasks."

He is right on both counts, and the second one is the interesting half. A plant does not become hardened off because you told an app it did.

What the phrase actually means

A seedling raised indoors has never been asked to do anything difficult. That is not its fault. The waxy layer on its leaves is thin, because in a warm room under a shop light nothing ever threatened to dry it out. Its stem is soft, because no wind ever pushed against it.

Put that plant outside for a full day and you are not introducing it to the sun. You are sandblasting it.

So you go slowly. An hour or two in full shade the first day, out of the wind — wind does more damage than cold at this stage, and almost nobody says so. A little more the next day. Some morning sun by the third. Most of a day by the fifth. One night left out at the end, and then it can go in the ground.

About a week. That is the whole technique, and it has been the same technique for a hundred years.

Two ways to lose the tray, and only one of them is ignorance

Skip hardening off entirely and you lose the seedlings to the first bright afternoon. That is a knowledge problem. An article fixes it.

The other one does not work like that.

You set the tray on the step at nine in the morning, feeling organized. Then the day happens — a phone call, somebody's kid, a job that runs long — and you remember at eleven at night, standing in the kitchen in your socks. They have been out for fourteen hours, through the whole temperature drop.

Nobody decides to do that. It is not a gap in what you know. It is a gap in what you were reminded of, and no amount of advice has ever closed it. Every guide on the internet explains the week correctly and then leaves you alone with it.

That is the part software can actually do something about, and it is the only reason this feature is worth building.

What our own data does not know

Before building the schedule we went looking for it in our plant database. We hold 151 crops, sourced from extension services, cited field by field.

38 of them list hardening off. Not one carries a duration.

The record is a single word — the name of the action — and nothing else. No hours. No number of days. No note about which crops need longer.

That gap is real and it matters, because a cabbage and a pepper do not want the same week. Brassicas take cold nights that would check a pepper hard. A tomato wants more sun-conditioning than a lettuce. The schedule we ship is the standard extension shape — about a week, starting in shade, building up — and it is the same one for all 38.

The app says so, on the card, where you can see it: this is general practice rather than a figure worked out for your crop. Sourcing it per crop family is on our list. Until it is done, the honest move is to tell you the schedule is generic rather than let you assume it was computed.

What we built instead of a button

Ten steps. Each one states the hours and where to stand the plant, and each one but the last ends with bringing it back in.

It was seven when this post first went up, and seven was wrong. We went back to the sources and got three answers, all of them longer than ours: Michigan State says begin seven to ten days before the frost-free date, Minnesota and Utah State both say about two weeks. Seven days is the short end of the range. We had been printing it as the range.

You tap when you put it out. The time is recorded, and the card turns into a countdown — out since 9:14, back in by 11:14. That single change is the whole point of the rebuild: the app now knows something it can remind you of, instead of knowing that you once pressed a button.

The rules that took the most thought

Two taps in one afternoon is one day. Bringing a tray in for lunch and putting it back out is normal, and so is a double tap. Counting taps would march somebody to "leave it out overnight" three days early, which is the exact failure the flow exists to prevent. Progress counts distinct days, and the hours accumulate within one.

A gap means backing up. Three days indoors and the plant has lost most of what it gained; the leaf conditioning reverses. So after a gap the schedule repeats the last step rather than advancing, and says why. Resuming at day five with an unconditioned plant is the scorch we were trying to avoid.

If nobody logs the tray coming back in, the app asks in the morning. You tapped it out at nine and then the log went quiet. Maybe you carried it in at noon and never said so. Maybe it sat on the step through the whole temperature drop. The app does not guess. The next morning it asks: did it come back in? The two possible answers are a shrug and an emergency, and it has no business picking one for you. Answer "it stayed out" and that gets recorded as a night out, with a note to check the leaves before carrying on.

It will not start you at eight in the evening. An early version cheerfully offered "2-3 hours outside, back in by 12:46 AM." The arithmetic was right and the instruction was nonsense. Sunlight is the point of the exercise, so a session that would finish after dark drops to the bottom of the list and says it is a morning job.

One card for the whole tray. Six plants take one button. They do not have to be on the same day of their own schedule — a tomato on day one and a pepper on day five sit together with their own numbers and their own bring-in times.

What it records

Each day keeps the time out, the time in, and how long that came to. Over a week that is a record of what happened rather than a row of ticks.

Which means the app can notice things. That a day ran three hours over. That a night got missed. That four days went by. None of that is available to a checkbox, and all of it changes what you should do next.

It also means the times have to be correctable. A timestamp you cannot fix is a timestamp you stop trusting, so tapping the recorded time lets you change it — "I actually put them out at eight" — and the countdown recalculates.

Where we are still wrong

The generic schedule turned out not to be the debt we called it. One shape, 38 crops — we went looking for the per-family version and there is no such thing. Michigan State, Minnesota and Utah State all describe the same ramp for a cabbage and a pepper. What actually varies is how much cold the plant will take before it is hurt, and that is already on the crop: 50°F for the tender ones, 40°F for the hardy. Two of the 38 run backwards. Celery must not be chilled at all. Artichoke has to be.

The in-app reminder alone does not reach a closed phone. It fires while the app is open or in the background, and it fires late the moment you next open the app if the time has passed. Getting it to a locked screen needs push notifications. Those are built, switched on, and one has arrived on a real phone with the app shut — 31 July 2026, reported by one person here, who also got a near-duplicate on opening the app that has since been fixed. That is one phone on one day rather than a test across devices, so send yourself a test alert and find out about yours.

The weather does not touch it yet. A frost warning appears above the hardening card, which is right, but the schedule itself does not know that tomorrow is 45°F and blowing. It should pause itself. It does not.

And the honest limit on all of it: a plant that has been through our seven days is not certified for anything. It is a plant that has had a reasonable week. Look at it before you put it in the ground.