Tesla’s Full Self-Driving (Supervised) software has made huge strides this year with its European expansion and the launch of commercial driverless Robotaxis, but even dedicated users know the system still has a few persistent blind spots. Two of the most common complaints among drivers center around route selection and rough pavement: navigation taking confusing detours, and FSD plowing straight through tire-popping potholes.
Fortunately, fixes for both headaches are on the horizon. In an exchange on X, Shopify executive Farhan Thawar praised FSD v14.3.10 but noted that broken asphalt and stubborn navigation routing remain his biggest friction points.
Responding directly, Phil Duan, a principal software engineer on Tesla’s AI team, confirmed that “improvements on both are coming.”
Improvements on both are coming.
— Phil Duan (@pduan) October 1, 2026
Pothole Detection
Dodging road craters has been one of the most requested capabilities among the fleet, and Tesla has been promising it since 2019. More recently, Elon Musk noted pothole avoidance is officially coming soon, and Tesla has actively listed it as an upcoming improvement in release notes since FSD v14.3 earlier this spring.
Current builds occasionally display glimpses of this behavior. One tester running FSD v14.2.2.5 previously recorded their vehicle gently centering over cracked asphalt so the crater passed safely between the tires. However, it’s not clear whether the neural net recognized it as a pothole or simply reacted to an ambiguous road anomaly, and the behavior has been inconsistent across recent FSD versions.
Once Tesla officially rolls out pothole avoidance, those rare maneuvers will turn into reliable, predictable driving habits. Rather than forcing drivers to manually disengage to protect their wheels and the occupants’ comfort, upcoming FSD models will detect broken surfaces in real time and steer smoothly around them.
FSD Navigation Improvements
Navigation logic is also receiving a massive overhaul. While FSD handles lane changes and turns well, drivers frequently complain about the overall route-picking logic and the system’s insistence on weird side-street detours instead of sticking to their regular commute.
Tesla is fixing this by teaching FSD to learn from your daily habits. By analyzing driver interventions, the system will use your manual lane choices and steering nudges as training feedback. If you consistently ignore a mapped detour to take a preferred highway exit or surface avenue, FSD will remember that preference on future trips rather than fighting the wheel. This capability will build upon the foundation Tesla recently laid with the new Preferred Routes feature.
What’s more, that learning will extend all the way to your destination. Upcoming FSD builds will study your parking habits, noting whether you prefer reversing into home driveways, hugging garage walls to leave walking room, or parking slightly further from high curbs. Once trained, the car will try to mimic those preferences when parking autonomously.
Other Upcoming FSD Features
Beyond pothole sensing and smarter routing, Tesla is working on integrating Grok directly into the FSD stack. Rather than tapping screen menus, drivers will soon be able to give FSD natural voice instructions — asking the car to take a specific turn, park in a certain spot, or change routes on the fly. Add to that the recently launched Automatic Collision Evasion feature that reengages FSD during manual driving, and Tesla is really cooking with FSD right now.
Pothole avoidance was initially promised in the v14.3 release notes, but it remains to be seen whether these upgrades arrive in an imminent point release or debut with the upcoming FSD v15 architecture, which is expected to bring a major neural network upgrade to 10 billion parameters. Either way, teaching the car to dodge craters and remember your favorite streets will make daily autonomous commuting far smoother.
