One Year In: What PlantLab Gets Right and Where It Still Falls Short
PlantLab started a little over a year ago, in October 2025, and went public that December. This post is a status report: what the system does today, which parts I trust, and which parts I do not.
What Is Live
Two plants: tomato and cannabis. Tomato went live this month.
Beyond naming a condition or a pest, PlantLab now does four more things:
- It reads the growth stage of a plant: seedling, vegetative, or flowering. Tomato does not have this yet.
- It sorts every finding into one of six coarse health groups, such as the mobile nutrient family. If the exact name is uncertain, the group is still useful.
- It reasons about likely root causes for some findings.
- It handles several plants in one photo. Each plant gets its own box and its own diagnosis.
What Works
| What | Result |
|---|---|
| Growth stage (seedling, vegetative, flowering) | about 90% |
| Healthy plant vs plant with a problem | about 80% |
| Potassium deficiency | about 80% |
| Nitrogen deficiency | about 80% |
| Magnesium deficiency | about 70% |
| Tomato: healthy plus nine conditions and pests, including septoria, spider mites, and late blight | about 90% or better each |
Tomato is the newer plant, so its numbers will move. I am still adding real-world tomato photos and training on them. They add the variation and adaptability that are a strength of the first plant’s model.
What Does Not Work As Well As It Should
Thrips, underwatering, and light burn are still performing poorly, and I am working on improving them. The same is true of zinc, manganese, and several pests.
Identifying individual plants is also still a difficult task, especially in dense canopies.
When PlantLab is unsure, it says so. A response with low confidence or a coarse group only is the system working as intended. It is not a bug.
Why the Second Plant Was Faster
The first plant took many months. The second took a small fraction of that time. Most of the work between the two went into making the process repeatable, so a new plant does not start from zero. The second plant is less polished than the first, and it is catching up.
What Is Next
Apple, grape, and leafy greens are the next plants on the roadmap. Between plants, there is a steady cycle of cleanup, research, and improvement that feeds every plant at once. More integrations also stay high on the list, because a diagnosis is more useful inside the tools growers already use. Home Assistant is the first of those, and the API documentation covers the rest.
Still Free
PlantLab stays free for now. I hope that lets growers of every kind try it, push on it, and tell me where it is wrong. If your photo gets a bad diagnosis, that is the most useful thing you can send me. You can try it at plantlab.ai.