← SnapRecaps

Tesla Self-Driving 14.3 - Is it ready for NYC?

► 4,952 views ⏲ 39:57 Watch on YouTube ↗

Summary

Tesla FSD 14.3.1 impresses in a chaotic NYC test with improved driving, but cut-ins, pothole issues, and a red-light violation show it's not ready for unsupervised use.

Executive Summary

In this hands-on review, the host finds Tesla FSD 14.3.1 “pretty good” and ready for wider release, praising its improved parking, decisive maneuvering, toll-booth selection, and pedestrian handling. The delayed rollout is likely due to bundling with a spring update. The ultimate test—a chaotic New York City commute—reveals that dense traffic isn't the real challenge; unpredictable conditions and last-second navigation changes are. Key limitations emerge: overly cautious low-speed following allows constant cut-ins, pothole avoidance is missing, and a red-arrow violation draws police attention. Still, the host stresses that FSD’s willingness to reroute rather than break laws is a major advantage, and he looks forward to fully unsupervised FSD, where drivers could sleep during commutes—a feature he considers worth top dollar.

Key Points

  • ▶ 0:53 FSD 14.3.1 has been “pretty good” with no major issues, offering an advantage over 14.2—especially in parking and decisive maneuvering—and the host believes it is ready to go wide.
  • ▶ 1:57 The delayed/limited rollout is likely because Tesla is bundling FSD 14.3 with its spring release, which most users will get as a combined update, possibly 14.3.2.
  • ▶ 4:55 FSD has done a good job selecting the correct toll booths, avoiding cash-only options and using E-ZPass booths, with no real issues encountered.
  • ▶ 10:32 Suggests a prerequisite for FSD on toll roads: the vehicle should have a toll tag (Easy Pass), since lane choice depends on knowing the car's payment capability.
  • ▶ 12:46 Notes a limitation: the visualization shows the car within the lane line, but in reality it crosses the line at times — so the display may not fully reflect actual positioning.
  • ▶ 15:53 Emphasizes that dense traffic is not the real challenge for FSD; unpredictable conditions like construction, closures, or last-second navigation changes requiring multi-lane crossings are.
  • ▶ 16:45 Driver observes how FSD handles approaching Park Street as part of a multi-stop route.
  • ▶ 16:54 Notes other drivers making illegal turns after getting caught in the wrong lane.
  • ▶ 17:07 Key contrast: FSD would not break traffic laws—it would reroute to the appropriate way instead.
  • ▶ 17:15 FSD successfully handles a dense, contested lane change, posturing and completing it smoothly despite close traffic.
  • ▶ 17:38 This commute is described as the ultimate FSD test and a top U.S. use case, with value hinging on freedom and stress reduction over pure economics.
  • ▶ 18:29 FSD drives assertively when changing multiple lanes but remains smooth and easy, though slower than a manual driver.
  • ▶ 18:47 FSD approaches a left turn in heavy traffic, searching for an opening in oncoming traffic while the driver notes it may wait for another driver or bypass the turn.
  • ▶ 19:00 FSD has two navigation options—turn left immediately or continue straight and turn later—but by 19:27 it's too late to get over, so FSD reverts to the alternate route.
  • ▶ 19:40 FSD commits to the maneuver and smoothly handles a truck pulling out, braking appropriately at 19:48, which the driver acknowledges at 19:52.
  • ▶ 20:01 FSD correctly identifies the proper turning lane in an ambiguous multi-lane intersection.
  • ▶ 20:06 FSD shows situational awareness by considering moving around a truck while accounting for oncoming cars.
  • ▶ 20:30 After checking the navigation display, the driver confirms FSD is positioned in the correct right turning lane.
  • ▶ 20:51 FSD needs assertiveness combined with an appropriate follow distance at low speeds for dense city driving.
  • ▶ 21:06 Allowing a single cut-off creates a compounding effect — multiple drivers cutting in can balloon a 20-minute trip to 23, 25, or longer.
  • ▶ 21:24 This assertiveness is critical in NYC because traffic lights don't last long; FSD must "get in quick" to complete turns before the light changes and avoid waiting another full cycle.
  • ▶ 21:43 The driver spots panhandlers approaching cars at the intersection and decides to ignore them as best he can.
  • ▶ 21:57 Heavy traffic and long red lights trap drivers in place, creating an ideal environment for people to pressure them for money.
  • ▶ 22:12 He grudgingly acknowledges the tactic is strategically effective given the traffic conditions.
  • ▶ 22:24 In heavy traffic, surrounding cars aggressively try to jump ahead of a truck, making FSD's right turn especially challenging.
  • ▶ 22:35 The driver deliberately avoids intervening (even with encouragement) to let FSD handle the situation hands-off.
  • ▶ 22:43 Core limitation: FSD's follow distance is too cautious at low speeds (below 50 mph), allowing other cars to cut in and preventing a smooth right turn.
  • ▶ 23:12 Driver declines a panhandler, setting the scene for chaotic urban traffic with cars cutting around a slow truck.
  • ▶ 23:29 FSD encounters a person in the road signaling to stop; despite confusion, it decides to go around and successfully executes the avoidance maneuver into a gap.
  • ▶ 23:59 Driver praises FSD's "good maneuver" but notes it "should have just stopped and waited," calling the assertive behavior "very Mad Max mode."
  • ▶ 24:26 Supervised FSD is very dangerous in heavy/stop-and-go traffic for drivers who are not well-rested or have sleep issues, because the relaxing, idle experience makes it easy to fall asleep; if FSD disengages, they'd be stranded in traffic.
  • ▶ 25:00 The host is excited about future unsupervised FSD, where drivers could nap and wake up at their destination—seen as super valuable and worth paying top dollar for.
  • ▶ 25:15 Unsupervised FSD would especially help daily commuters and those driving at odd hours (middle of night, very early or late), enabling sleep during the commute.
  • ▶ 25:34 The drive is roughly 42 minutes into the actual journey.
  • [25:37-25:41] The narrator fast-forwarded through the tunnel traffic, so the on-screen elapsed time is shorter than the real drive time.
  • ▶ 25:41 The segment resets viewer expectations by noting the actual real-world duration (though the thought is cut off).
  • ▶ 25:58 FSD stops blocking a crosswalk, which the driver immediately flags as "not great."
  • ▶ 26:02 A Mercedes EQE cuts in front of the Tesla because FSD is not assertive enough to complete the maneuver.
  • ▶ 26:35 FSD violates a red arrow (with no "No Turn on Red" sign), drawing police attention, and the driver jokes, "Officer, car did it. It wasn’t me."
  • ▶ 31:31 FSD shows clear pedestrian-handling improvement over 14.2, avoiding overreaction to pedestrians not in its path while still stopping appropriately for those who might step out.
  • ▶ 34:16 Pothole avoidance is a major gap: the car hits a bad pothole and the driver says “We need pothole avoidance yesterday,” noting it could have easily avoided it.
  • ▶ 37:49 A post-stop bug triggers a brief “auto steer unavailable” message after rapid stop-to-stop successions, but FSD re-engages immediately—flagged as a bug to report rather than a disengagement.

Video Sections

  • ▶ 0:00 Introduction, FSD Performance, and Toll Booth Approach (0:00 - 5:43) - Welcome, initial FSD impressions, release notes, Mad Max mode, route check, and toll booth performance.
  • ▶ 5:43 Traffic, Tunnel Approach, and NYC Driving (5:43 - 16:54) - Clearing traffic, NYC view, toll and tunnel lane choices, entering the tunnel, and dense traffic reflections.
  • ▶ 16:54 Heavy Traffic Maneuvers, Assertiveness, and FSD Fatigue (16:54 - 27:01) - FSD handles illegal turns, heavy traffic, left turns, panhandlers, assertive two-lane moves, and fatigue risks.
  • ▶ 27:01 Construction, Pedestrians, and Final Arrival (27:01 - 39:58) - Soul-crushing traffic, tight spaces, pedestrian and cyclist responses, construction zones, and arrival in NYC.

Exact Transcript

Load the full timestamped transcript on demand and click any time to jump in the video.