I'm Building Tomato Diagnosis Next

Tomato diagnosis is now available. I’ve picked the second crop, and it’s tomato: the most-grown, most-photographed, most-argued-about plant in the garden and the field. Starting now, I’m building tomato diagnosis the same way I built the first one.

The problems between here and a tomato model

The lookalike exam. Early blight, septoria, and late blight read almost alike at the stage where catching them matters most - a pinhead spot on an oldest leaf. Telling them apart from a photo, early, is the whole job. Everything else is plumbing.

Real photos, not textbook photos. Field tomatoes grow in crowded, backlit canopies with dirt-splashed lower leaves. Textbook single-leaf shots are easy. The photos growers actually take are the ones that have to work.

Trust is per-crop. Whatever the lab proved on its first crop, tomato doesn’t inherit it. The model earns trust in the field, against the hard cases, or it doesn’t earn it at all. Starting from zero is the price of doing it right.

It ships when it clears the same bar the live crop cleared, and I’ll post the real numbers as they come.

Watch it happen at plantlab.ai/platform, or get notified there when tomato’s status changes.