A farmer with two hectares knows his soil better than any sensor will. He has walked it for thirty years. He knows the corner that floods and the corner that burns.
What he does not have is the thing a large operation takes for granted: a number he can act on before the season goes wrong. Big agriculture has soil labs, agronomists on retainer, and satellite imagery. A smallholder has his eyes, his memory, and a guess about fertiliser.
That gap is not a technology problem. The technology exists. It is a distribution problem — the tools were simply never built for him.
Terrafarms was our attempt at closing a bit of it. It came out of Bangkit, Google's academy here in Indonesia, and it pairs AI with IoT sensors to do the boring, expensive things: analyse the soil, recommend what to plant, catch plant disease early from a photo, and monitor nutrition in real time. It also tries to solve the other half of the problem nobody demos on stage — actually getting hold of decent seed and fertiliser, and actually finding a buyer.
I want to be careful not to oversell this. A dashboard does not fix a broken supply chain, and an app does not make it rain. The hardest problems in Indonesian agriculture are not machine-learning problems; they are road problems, credit problems, and price problems. Anyone who tells you otherwise is selling something.
But we cannot fix everything is a terrible reason to fix nothing. If a farmer catches a disease a week earlier because he photographed a leaf, that week is real. It is his margin.
That was the whole idea.
