Since Tesla introduced its Full Self-Driving system, unprotected left turns have proven to be one of the hardest problems for the software to manage. For over three years, Tesla’s engineers have collected data and made improvements to tackle this challenge. The company officially acknowledged the difficulty in 2022 after Chuck Cook demonstrated FSD v10’s struggles at a Jacksonville, Florida intersection near Memorial Park.
Early versions of FSD required simultaneous clear gaps on both left and right traffic lanes before attempting a turn. The system moved slowly when small gaps appeared in traffic. It couldn’t effectively employ medians for safer turning patterns. The algorithm simply checked if conditions were safe rather than predicting what would happen next.
Tesla documented safety concerns in FSD v13. The system accelerated inappropriately during left turn initiation, even when oncoming traffic traveled at high speeds in lower-speed zones. Near-miss incidents occurred during turns involving chicanes and tram line interactions. These reproducible safety risks showed the system wasn’t ready for real-world complexity.
However, FSD v14.1.2 showed dramatic improvements. The system now steers through the Jacksonville intersection smoothly, even during low-light conditions with headlight glare. It successfully employs medians and handles small safe gaps without excessive hesitation. The software now anticipates traffic patterns instead of simply reacting to current conditions.
Tesla has identified at least 1,000 similar unprotected left turns across the United States. The company officially named the Jacksonville test location the “Chuck Cook-style” intersection in FSD Beta v10.13 release notes. Tesla employees regularly perform validation tests there during different lighting conditions, including early morning twilight around 7:15 AM.
The evolution from FSD v10 through v14.1.2 shows steady progress. Version 12.4.1 included nine documented attempt sequences at the challenging Jacksonville location. Each new version brought incremental improvements in decision-making and execution.
Tesla’s approach demonstrates how difficult autonomous driving really is. What humans do instinctively—turning left across traffic—requires sophisticated software to manage safely and reliably. The improvements in recent versions suggest Tesla’s engineers are making real progress on one of self-driving’s toughest challenges. Tesla’s continuous data collection from nine billion miles of real-world driving helps refine these complex maneuvers and improve safety performance across all autonomous driving scenarios.