After years spent scaling storage, databases and machine learning, the Seattle engineer found his most demanding computer at ground level - where dust, weather and skeptical farmers make every elegant idea prove itself.

After Mars-yard robots, self-driving cars and sidewalk couriers, Gabe Sibley found a more immediate autonomy problem in a farm field - one weed, one nozzle and one careful shot at a time.

He started with neural prosthetics, followed a water problem into agriculture, and discovered that the hardest part of a tomato is persuading a machine to pick it quickly, gently and all day long.