Mark Rober’s now infamous Tesla “wall test” sparked a lot of debate around the capabilities of Full Self-Driving (FSD). While Rober ran the experiment using a Model Y on Autopilot, and not on FSD, YouTuber Kyle Paul has recreated the test using two vehicles, this time on FSD Hardware 3 (HW3) and the newer Hardware 4 (HW4), and actually using FSD.
Following the backlash on his misleading test, Rober dismissed the difference between Autopilot and FSD as a “distinction without a difference.” However, Paul’s results challenge that assumption, as the recreated test ends with different, but also similar results.
Paul used a 2022 Tesla Model Y equipped with HW3 and running FSD version 12.5.4.2. The setup itself mirrored Rober’s original test, which placed a printed road scene on a wall in the vehicle’s path, designed to trick the system into thinking the road continued. In the recreated test, the Model Y failed to recognize the wall in time, with the driver having to intervene manually to prevent a collision.
Switching to a Cybertruck equipped with HW4 and the latest FSD version 13.2.8, the outcome was different. The Cybertruck recognized the wall from a safe distance and came to a full stop on its own, without any driver input, in contrast to the Model Y’s failure.
But, speaking of contrast, the test conditions were not identical. The Cybertruck’s test was conducted later in the day, near dusk, when lighting and environmental contrast were more pronounced. As you can see in the images above, the wall, essentially a visual illusion, stood out more clearly against its surroundings in the Cybertruck test, potentially giving it an edge over the Model Y. This introduces the possibility that environmental factors, in addition to hardware and software, may have played a role in the improved result.
Either way, it shows that Rober’s initial test was flawed, and that it should be redone under the same conditions using HW4 and the latest version of FSD.
You can watch the full test below.
