How it works
The logic behind every reminder.
No magic, no black box. Here's exactly how PLAN'T decides when and how much to water, using the same calculations running in your dashboard right now.
Curated species
112
Values behind each date
12
Pot sizes, each its own pour
4
Botanical frames per reading
Up to 4
Identification
One photo guesses. A few photos know.
Plant identification lives or dies on the photo. In our own testing, a single photo of a golden pothos scored the right species at just 17.8% confidence. Ranked second, behind a visually similar wrong match. Two photos of the same plant scored 96.3%. That's why PLAN'T asks for a leaf, the whole plant, and a close-up, not just one snapshot.
One plant, one test of our own, not a measured accuracy rate
Scheduling
Not a generic schedule. Your schedule.
Every species has a baseline watering interval, and eleven things move it. The weather where your plant lives comes first: heat shortens the interval, cold lengthens it, and dry air shortens it too. Measured as vapour pressure deficit, because thirty percent humidity at thirty degrees is not thirty percent at ten. Indoors the swing is halved, since climate control smooths out extremes a balcony gets in full. Then the physical facts: the vessel, the mix inside it, whether the pot can drain, and how much of it the roots have taken. Then the length of the day from your latitude, and rain for anything kept outside. Last comes what you found the previous times you put a finger in the soil. The only term with evidence rather than consensus behind it, which is why it goes last and absorbs what the others miss.
Next watering in 7 days
This is the actual scheduling function running live. Move the controls.
Amount
How much to water isn't about the plant. It's about the pot.
The amount of water needed to fully saturate the soil is mostly a function of pot volume, not species. Species mainly changes how often you water, not how much each time. Heat, dryness, and sun exposure nudge it a little further on top of that. Tap through the sizes below.
The exact figures used across your dashboard, for a pot that drains. In a pot with no hole the pour is cut to a third, because water that cannot leave becomes a reservoir at the roots.
Diagnosis
One leaf tells on the whole plant.
Photograph a worried leaf and three readings meet. Pl@ntNet's disease classifier (a real model over a closed set of coded diseases) supplies the ranked causes. A vision model describes what is actually visible and writes the steps, because that is the one thing a classifier cannot do. And your own watering log settles the question a photograph never can: over- and under-watering look identical on a leaf, so the log can promote a cause or overrule it. Each cause carries a band rather than a percentage, with the source named. And a reading the vision model reached alone can never read stronger than "possible", because an uncalibrated self-report has not earned a stronger word. It will also happily tell you the plant is fine.
Leaf scorch
Light
What to do
Move it a step back from the south glass.
Water in the morning, before the heat.
An illustrative reading, run a real one on the Diagnose page.