How we work, and how we'll prove it.
LeafShift is in active research and development. We build carefully, on established science, and we hold ourselves to one rule: we do not claim what we cannot show. This page is how we think about proof. The results, and their limits, follow as our pilots run.
Our decisions are built on FAO-56, the international standard for crop water use, and on decades of peer-reviewed hydrology. We run on established, openly validated implementations of that science. We are a research-minded team, and the method reflects it.
A model that looks convincing is not proof that it is right. Before we make any claim, the model has to earn it, one step at a time.
Check the model against reality
Before any claim, does our water balance track real, measured conditions on the ground? If it does not, nothing built on top of it counts.
Test the decision it produces
Would our choices actually have changed the outcome on a real season, holding everything else fixed? A model that fits the data still has to prove it makes a better decision.
Hold it across many seasons
Does the advantage survive years of real weather and honest uncertainty, not one convenient season? Only then does it earn a general statement.
A later step does not count until the one before it holds. That order is the whole discipline.
The core of LeafShift is a physical water balance you can follow line by line. Machine learning is added only where it measurably improves a decision, and always within bounds we can check. Every step of the reasoning stays visible.
When we report a result, it will come with what it does and does not show: the crop, the place, the conditions, and the caveats. We will tell you where we are still uncertain. Being clear about limits is central to how we work.
The real proof comes from the field.
Everything so far is careful modelling and method. The evidence that matters is measured on real ground: LeafShift running alongside a real operation, logging what it would have done next to what actually happened, over a real season, with the grower's own records. That is what we are building toward, and what we will publish, results and limits together.
The research log.
Occasional notes on how the work is going, by email. No noise, and you can unsubscribe from any message.
Built by a small, research-background team. Meet the team.