A roboticist built a hardware aimbot that could outperform the pros, until it aimed so hard it died

pcgamer.com:

First spotted by Hackaday, an industrious robot-builder, Kamal Carter, has created a physical aimbot that will visually scan a computer screen and then physically move a mouse to click on targets, and it's good enough that it can outperform some Valorant pros in aim training software. Or could, before its brief shot at esports glory was snuffed out.

Aimbotting is more typically accomplished through software, removing our unreliable meatspace reflexes from the equation so we can click on heads with ruthless machine precision. It's a great bugbear of competitive FPS games, with its alleged use by opponents being second only to FPS players' own teammates as their most-cited reason for losing a game. Cheating software can be a real problem, pervasive enough that developers invest in anticheat solutions or expensive legal campaigns against their creators in order to preserve the competitive integrity of their games. 

To make his physical aimbot, Carter designed a chassis with four omni-directional wheels made to fit around a wireless mouse. This housing takes instruction from a program which can parse visual data, allowing the physical aimbot to react to events onscreen much as a human would.

Carter tested the device on an aim training program called Aim Lab which provides an objective measure of its effectiveness, as well as distinct targets in a sparse environment to calibrate the program on. Over two months of work, Carter got the aimbot to quickly and smoothly track targets without overshooting.

An average joe can expect an Aim Lab score of 40-50,000, while professional FPS players might secure one in the 80-90,000 range. Carter was able to achieve an Aim Lab high score of 118,494 with the robot. He was hoping to develop