This Crazy Aimbot Is a Literal Aiming Robot

When people throw around the term “aimbot,” they’re not talking about an actual robot sitting at a desk working the WASD keys and mouse. They’re talking about special software that uses information from the game (possibly analyzing the screen, but usually through more direct means) and then moves the aiming reticle (again, in software) to shoot opponents more quickly than a human ever could. As you would expect, game developers do their best to prevent that kind of cheating. To circumvent those efforts, Kamal Carter built a literal aiming robot.

Anti-cheat measures vary in sophistication and have evolved a lot of the years. They tend to be particularly advanced in highly competitive online shooters that have rankings or tournaments in which real-world money might be involved. On the client side, they can look for software that emulates or otherwise manipulates the inputs. On the server side, they might look for aberrant behavior — something that is often circumstantial and therefore less reliable.

In this case, Carter definitely bypassed any of that client-side detection, because his computer doesn’t need to run any input-manipulation software. Server-side detection might still be possible, if it can detect aiming performance that a human could never achieve. But that’s a long shot and Carter is probably safe.

Carter’s aiming robot works by looking at the screen and using a YOLO machine learning model to detect opponents and their positions. It then physically moves the mouse and performs a mouse click when the reticle is over an enemy. It clicks the mouse electrically using a relay.

Except it isn’t actually moving the mouse. The mouse is stationary and the robot (an inexpensive repurposed desktop CNC router) moves a table underneath. But the effect is the same, as the relative motion between the mouse and the table is all that matters.

Carter doesn’t say exactly what hardware he’s using to control the CNC router or to activate the relay. But he’s running the YOLO model on a gaming PC with a powerful GPU and it is probably interfacing with the relay through a microcontroller development board.

That worked well against the easy bots in Valorant, but as the difficulty ramped up, Carter found that the performance wasn’t quite up to snuff. An upgrade to better actuators on the CNC router and a better GPU on the PC solved that problem and by the end of his video, Carter was annihilating even the hardest in-game bots at rates better than just about any human could achieve — but not quite good enough to raise red flags with the anti-cheat software.