The video documents a series of real‑world tests of Tesla’s Full Self‑Driving (FSD) system, contrasting the capabilities of older Hardware 3 vehicles (e.g., a 2022 Model X) with newer Hardware 4 models (e.g., a Model Y). The creator first shows the Model Y successfully reversing into a tight garage, mimicking the driver’s habitual backing‑in behavior and adjusting its angle to avoid obstacles—a feat that seemed impossible a few months earlier. When the same test is run on the Hardware 3 Model X, the car struggles: it hesitates, fails to reverse properly, sometimes drifts toward a neighbor’s driveway, and appears unable to execute the maneuver without human intervention. The discussion highlights that Hardware 4 provides newer cameras, more compute headroom, and additional software features (such as destination‑based parking options) that Hardware 3 lacks, leading to a noticeable performance gap. The creator also explores how environmental factors (rain, construction, regional data laws) affect FSD learning, clarifies that personal FSD‑equipped Teslas remain driver‑supervised while a future Robotaxi service would operate in a tightly controlled environment, and notes ongoing debates about safety metrics and insurance implications. Overall, the footage illustrates rapid progress in Tesla’s vision‑only, neural‑net‑driven autonomy, while underscoring that hardware limitations still create divergent experiences across the fleet.
1. The Tesla reversed itself in full self‑driving into a garage around another car with very little clearance.
2. After the video was posted online, viewers asked questions about the maneuver.
3. The car’s display indicated degraded weather due to rain.
4. PJ was present with a boot on his foot while the car was driving.
5. The car angled itself to back in, mimicking behavior observed from the driver’s prior actions.
6. The car used its known dimensions to avoid clipping the garage.
7. The car backed in at an angle to avoid hitting the garage side.
8. A few months prior, the car was not capable of performing this maneuver.
9. The driver wondered whether Tesla’s hardware 4 performs better than hardware 3 in this task.
10. The driver questioned whether rain or sunlight affects camera performance.
11. The driver questioned what would happen if more obstacles were added.
12. The driver questioned why Robo Taxi considers a different approach if Tesla’s system is improving.
13. Many people believe Tesla FSD is limited to highway lane centering, but the car can now start navigation from park, reposition, reverse into spaces, and handle complex environments.
14. Tesla’s system uses cameras and a neural net, not LIDAR, to interpret surroundings.
15. Not all Tesla vehicles improve at the same rate with FSD updates.
16. Vehicles purchased with FSD years ago may have older hardware that limits future capability.
17. The 2022 Model X is equipped with hardware 3; the Model Y tested has hardware 4.
18. Hardware 4 provides newer cameras and greater computational capacity than hardware 3.
19. Hardware 3 vehicles can still run FSD under supervision.
20. Some Tesla owners question whether hardware 3 and hardware 4 platforms are improving at the same pace.
21. FSD capability depends on both software and the underlying hardware in the vehicle.
22. Producing and editing the video required tens of hours of work.
23. The video’s sponsor, Monarch Money, aggregates bank accounts, loans, investments, and credit cards into a single dashboard.
24. Monarch Money provides a financial overview in approximately thirty seconds.
25. Monarch Money reduces the need to switch between multiple apps or spreadsheets.
26. Monarch Money shows upcoming credit‑card payments, spending categories, and financial trends.
27. Monarch Money offers AI‑generated insights explained in plain language.
28. Monarch Money does not sell user data and is ad‑free.
29. Monarch Money connects to over 13,000 financial institutions.
30. Monarch Money uses bank‑level encryption and provides read‑only access to accounts.
31. The Model X was running FSD version 12.6.4; the Model Y was running version 14.3.3.3.
32. Hardware 4 enables initiating FSD from a complete stop.
33. The Model Y began performing the reverse‑into‑garage maneuver a couple of weeks prior to the test.
34. The Model X exhibited more steering‑wheel jerk and brake pumping than the Model Y during the test.
35. The Model X appeared to want to turn right, whereas the Model Y tended to back in.
36. Commenters noted that hardware 3 cannot reverse in full self‑driving mode.
37. During a test, the hardware 3 vehicle entered a neighbor’s driveway instead of the garage.
38. The hardware 3 vehicle remained engaged in FSD for several minutes without completing the maneuver.
39. Tesla is developing a reduced‑feature version of hardware 4 capabilities for hardware 3 vehicles.
40. Some hardware 3 vehicles lack a forward‑facing camera, while others include one.
41. Hardware 2 and hardware 1 vehicles exist in the Tesla fleet.
42. Moving the Model Y to the street removed obstacles from the driveway path.
43. With the Model Y on the street, the system at least attempted to pull into the driveway.
44. The Model Y attempted to back into the garage during the test.
45. Owners who have not upgraded hardware in the last year and a half report a noticeable difference between hardware 3 and hardware 4 performance.
46. Hardware 4 vehicles can detect a closed gate, slow down, and wait for it to open, rather than driving through.
47. In rain, a hardware 4 vehicle displayed a “FSD supervised may be degraded” warning but continued to drive.
48. Construction was present on the road during the hardware 4 test.
49. Tesla states that its telemetry data excludes certain regions (e.g., China) due to local data laws.
50. The usefulness of a mile of driving data varies by region because of differing regulations and conditions.
51. Europe imposes major regulatory restrictions on advanced autonomous features.
52. FSD behavior can differ depending on the geographic location of the vehicle.
53. A personal Tesla with FSD remains the driver’s legal responsibility; the driver must supervise and can intervene.
54. Drivers are expected to pay attention at all times when using FSD on a personal vehicle.
55. Robo Taxi service is operated by Tesla, which defines the service area, operating conditions, deployment zones, and permissible environments.
56. Unlike a personal Tesla, a Robo Taxi only needs to master a limited, controlled environment.
57. The current FSD version allows the driver to briefly glance at a phone without triggering a warning, under certain safe conditions.
58. Tesla aims to eliminate the driver‑monitoring system within the next one to two years.
59. Recent FSD updates have reduced the frequency of driver‑alert prompts when conditions are deemed safe.
60. Earlier FSD versions would immediately flag any glance at a phone.
61. The driver considered the garage maneuver observed to be potentially more challenging than some Robo Taxi scenarios.
62. The driver planned an additional test involving a bicycle placed in the garage to assess obstacle avoidance.
63. The test setup created an obstacle course with the bicycle and a parked car inside the garage.
64. The Model Y provides destination‑behavior options (e.g., park at charger, parking lot, street, or curb) that the Model X does not.
65. The Model Y’s interface includes a charger tab that, when accessible, lets the driver select a charging destination.
66. The charger option appears only if the vehicle’s location history shows a home charger.
67. When the vehicle detects a known home charger location, it automatically offers to park there.
68. Destination‑behavior selection is unavailable on older Tesla hardware.
69. Tesla introduced a public statistic showing the user’s FSD streak (consecutive miles driven with FSD engaged).
70. Driving with FSD is no longer described by the speaker as an expensive gamble.
71. Insurance companies can compare miles driven with FSD supervision to miles driven manually to assess risk.
72. Tesla’s insurance product can incorporate the proportion of FSD‑supervised miles when calculating certain premiums.
73. Lemonade has launched insurance offerings that integrate directly with Tesla vehicle data.
74. During a later test, the vehicle chose to drive forward around a bicycle instead of reversing.
75. The bicycle’s small size allowed the vehicle to navigate past it without needing to back up.
76. The vehicle’s performance was slower in the second attempt compared to the first.
77. The driver disengaged the system after perceiving a risk of collision.
78. The system showed steering‑wheel adjustment before the driver intervened.
79. The driver believed a collision would have occurred if the vehicle had continued straight backward without steering correction.
80. Tesla asserts that supervised FSD reduces collision rates compared to human driving.
81. Opinions differ on whether FSD is safer than a human driver, with conflicting views online.
82. Tesla’s method of counting collisions differs from some regulatory approaches, complicating direct comparisons.
83. Tesla counts miles driven while FSD is engaged; there is debate about whether miles in closed‑loop test environments should be included.
84. Ten years ago, the maneuvers shown in the video would not have been possible with available technology.
85. Current discussion focuses on whether the technology is sufficiently reliable for broader use.