How GrowCast thinks
A decision engine, not another weather summary.
We translate forecast data into reversible, risk-weighted garden actions. The aim is not false certainty. It is a better next move.
1. We score actions, not weather
Pleasant weather and useful working weather are different. Each action has its own failure modes: wind creates spray drift, late rain shortens product contact time, and heat raises transplant risk. One universal “garden score” would hide those differences, so the action cards remain primary.
2. The cost of being wrong changes the threshold
A false “water” may waste water. A false “safe to spray” can waste product, damage non-target plants and leave the crop unprotected. Higher-consequence actions require a wider safety margin.
3. Near the threshold, observation beats confidence
Forecasts cannot see the moisture two inches below your mulch or the wind shadow behind a fence. When the model sits near a decision boundary, GrowCast asks for the smallest physical check capable of changing the answer.
4. Current limitations are part of the product
The initial model uses National Weather Service forecast data and ZIP-level location. It does not yet know your soil texture, irrigation history, slope, crop stage or microclimate. Recommendations are decision support and should be corrected by direct observation and product labels.