The question
Every VEX match opens with fifteen seconds where the robot drives itself, then hands over to the drivers. So every team ends up asking the same thing: is our autonomous routine actually better than driving it ourselves? Then AI tools started helping teams write their routines, and the question got bigger. Is the code an AI wrote better than ours? Better than a good driver?
Nobody could answer it with a number. A match score mixes in the other alliance, the field, and luck. Two teams can’t compare their runs, and nobody measures the drivers at all.
What they built
So they set out to measure it properly: the same tasks on the same field, every run recorded the same way, and every result compared with real people doing the same job. Once that worked on a VEX field, they noticed it worked anywhere a machine moves around a bounded space with a goal: a robot vacuum in a living room, a mower in a yard, a robot arm on a classroom bench.
That became MotionBench: open source, cheap enough that any team with a phone can take part, and honest about what the numbers show. You can read the full commitments in why we’re doing this.