FSD in the endurance test: Tesla Cybertruck and Model Y in the "Wall T

After the much-discussed "Wall Test" by Mark Rober, a new video provides exciting insights into the performance of Tesla's Full Self-Driving. This time the test was conducted under real conditions with FSD (not just Autopilot) – and with two different vehicles: a Tesla Model Y with FSD Hardware 3 (HW3) and a Cybertruck with the new Hardware 4 (HW4). The results? Different – and yet similar.

Reminder: What is the "Wall Test"?

The original test by Mark Rober caused a stir: A printed road motif was stuck on a wall to deceive the Tesla into thinking there was a continuing road. The question was: Does the vehicle recognize the wall as an obstacle – or does it blindly continue?

Rober used Autopilot for this, not FSD, which led to criticism. Nevertheless, he later described the difference as "insignificant" – an assessment that is now being questioned.

The new test: HW3 vs. HW4 in direct comparison

YouTuber Kyle Paul has re-conducted the test – this time correctly with activated Full Self-Driving, and in two variants:

Model Y (2022) with HW3 and FSD Version 12.5.4.2
Cybertruck with HW4 and FSD Version 13.2.8

Both vehicles were subjected to exactly the same setup – an optical illusion in the form of a painted road leading against a wall.

Result:

The Model Y with HW3 could not recognize the obstacle in time. The driver had to intervene to prevent a collision.
The Cybertruck with HW4 stopped independently at a safe distance – without intervention.

A clear plus for the new hardware generation.

But it's not that simple: The conditions were different

An important point: The Cybertruck test took place at dusk, when the lighting conditions and contrast of the wall were significantly better than with the Model Y. This made the visual illusion easier to recognize – also for the camera system.

This means: Not only software and hardware, but also environmental factors play a crucial role.

What does this tell us?

FSD is constantly evolving. The newer versions and hardware generations offer clear advantages in object recognition and situation analysis.
Test conditions must be identical to be able to draw reliable conclusions. The comparison of HW3 and HW4 shows progress, but no complete failure of the older hardware.
Rober's criticism is not unfounded – but abbreviated. The difference between Autopilot and FSD is not trivial, as this test shows.

Conclusion: Progress visible, but not yet perfect

The Cybertruck with FSD HW4 passed the test, where the Model Y with HW3 still failed. However, the different lighting conditions put the clear difference into perspective. What remains is a clear sign: FSD is getting better – step by step.

For Tesla drivers, this means: Those who drive newer vehicles benefit noticeably from better assistance systems. And those who want to seriously test FSD should pay attention to identical conditions – because only then can real progress be evaluated.