Kyle Vogt: Cruise Automation

Vogt traces his interest in robotics to an unusually hands-on form of learning. Lego Mindstorms made programmable machines tangible, while BattleBots supplied a concrete engineering challenge: he read thousands of forum posts, inferred how experienced builders combined components, and built wedge-shaped robots as a teenager. The attraction was not only spectacle. Running motors at two or three times their rated voltage, accepting overheating, and observing where batteries and materials failed made the competition a laboratory for limits, trade-offs, and rapid iteration.

The same curiosity moved from hardware to software. His first programming experiences involved Apple II machines and BASIC, where even inserting a line of code required leaving room between line numbers. A long, boring drive across Kansas then reframed autonomous driving as an approachable problem: detect yellow lane markers and steer. Vogt emphasizes that his early system did not depend on modern neural networks. On a 233 MHz Pentium, simple color thresholds and geometric heuristics were realistic tools, and at Cruise the team initially used human-written scaffolding before migrating particular capabilities toward deep learning.