Tesla's Full Self-Driving (Supervised) software is building a reputation for processing real-world hazards faster than a human ever could. To find out exactly how quickly the neural networks can compute a surprise obstacle, longtime Tesla enthusiasts AI DRIVR and David Moss recently put FSD’s safety logic through its paces in a new YouTube video.
Moss previously made headlines in the community for reaching almost 13,000 miles of intervention-free FSD driving and completing the first fully autonomous, coast-to-coast FSD drive across Canada without human input. For this experiment, the duo used a life-size blow-up doll they lovingly named Dan O as a crash test dummy, intentionally tossing him directly into the path of a Model Y being driven by FSD (Supervised) at various speeds and tightly timed intervals to evaluate the system’s reaction window.
Blinking Faster Than Humanly Possible
AI DRIVR touted FSD v14.3 as having an “inhuman reaction time.” To prove it, they used a roof-mounted camera to frame the exact moment Dan O's body edged past the curb and into the Model Y’s path.
At 32 mph, the FSD visualizer tracked the obstacle and initiated an aggressive evasive maneuver in just four frames. Running at 30 frames per second, that comes out to a split-second reaction time of just 0.13 seconds — quite literally faster than most people can blink their eyes. Even when Moss threw the dummy straight on the asphalt at 40 mph to simulate a pedestrian tripping forward, the car performed a picture-perfect swerve to clear the target.
Elon Musk has long maintained that the current v14 stack is already far safer than the average driver, and showings like this one reinforce that narrative. Tesla is currently planning to launch FSD version 15 toward the end of this year or early next year, which is expected to introduce a massive 10x parameter upgrade to increase this safety gap by orders of magnitude.
The Blind Spot Dilemma
The testing did expose a fascinating scenario when the car managed to stop and avoid the dummy, but then decided to shift into reverse. Because older (pre-Juniper refresh) Model Ys don’t have a front bumper camera, backing up placed the fallen dummy directly into a blind spot. The center screen flashed an "Object in path detected" message, but because of the lack of a low-level front camera, FSD appeared to have lost track of the dummy in its path and tried to pull forward again. Check out the full obstacle avoidance video above to view every high-speed test run.
Update 2026.14.6.10
Percent of Fleet
Installs on Jul 26
Tesla originally began rolling out FSD v14.3 earlier this spring with software update 2026.2.9.6, and point releases have come out in quick succession since. Just this week, software update 2026.14.6.10 started hitting vehicles, introducing new FSD streak celebrations and finally dropping Actually Smart Summon for the Cybertruck.
