What AI Plant Diagnosis Will Look Like for Tomatoes
Tomato is the crop I get asked about most after cannabis. The honest answer first: tomato diagnosis does not exist at PlantLab yet. Cannabis is the only live model. Tomato is the first crop on the public roadmap, and its status today is planned – not in development, not in beta, no ship date.
The short version
- Not live. PlantLab does not diagnose tomato plants today. Anything that claims to is not us.
- First in line. Tomato leads the public roadmap, ahead of apple.
- Planned, honestly. “Planned” means intended, with no corpus work started yet. I won't call it “in development” until that's true.
- Available now: cannabis diagnosis, free tier, 3 checks a day, at plantlab.ai.
The tomato scouting problem
Anyone who grows tomatoes at scale knows the routine: walk the rows, flip leaves, look for the first spot. It's hot, repetitive work, and the difference between a bad week and a lost season is usually one leaf noticed two days earlier.
The problems a tomato diagnosis model has to catch are public agronomy, documented anywhere tomatoes are studied:
- Early blight (Alternaria solani). Brown spots with concentric target rings, oldest leaves first. Common everywhere tomatoes grow.
- Septoria leaf spot. Many small spots, pale centers, dark edges. Starts low and climbs the canopy.
- Late blight (Phytophthora infestans). Greasy dark lesions on leaves and stems that can kill a planting in days. The one everyone fears.
- Leaf mold (Passalora fulva). Velvety growth on leaf undersides, mainly in greenhouses and high tunnels.
- Tomato spotted wilt virus (TSWV). Bronzed, ring-spotted young leaves. Thrips-vectored, no cure – removal is the only answer.
- Two-spotted spider mites. The same pest cannabis growers fight, same stippling, same fine webbing.
- Nutrient deficiencies. Nitrogen and magnesium show on the oldest leaves first, potassium scorches leaf margins – the same mobile-nutrient grammar across crops.
None of this is exotic. The hard part isn't knowing the list. It's telling them apart from a photo, early, when the spot is pinhead-sized and three of these look nearly identical.
What transfers from cannabis, and what does not
PlantLab's habits transfer: every result carries a condition, a calibrated confidence, and a severity reading, and the model is built to flag the mild early signal a person would skim past. A calibrated hedge beats a confident wrong answer, on any crop.
The lookalike knowledge does not transfer. Early blight versus septoria versus late blight on a stressed plant is its own exam, and the only way to pass it is the same standard cannabis met before its model shipped. No shortcuts, and I won't pretend otherwise.
What “planned” means
The crop roadmap uses four labels, and each one is a claim:
- Live – a deployed model diagnosing that crop. Today that's cannabis alone.
- In development – corpus work has genuinely started. Tomato isn't here yet.
- Research – an active, named collaboration or dataset evaluation.
- Planned – intended, no work started. Tomato sits here.
A crop moves up only when the evidence moves. No dates, no percentages, no “coming soon” until a model actually clears the bar.
What you can do today
If you grow tomatoes, the crop roadmap page has the current status and a notify option: plantlab.ai/platform.
If you grow cannabis, diagnosis works today – free tier, 3 checks a day, structured output through the web app or API.
Tomato diagnosis is planned, not live. Cannabis diagnosis works today at plantlab.ai – free tier, 3 checks a day. Watch the crop roadmap for tomato.