Merritt: Tesla FSD v14.3.1 Makes 'Mad Max Mode' More Controlled and Plans Lane Changes in Advance on Highways
Tesla autonomous driving practitioner Sawyer Merritt pointed out on his personal platform that after enabling FSD v14.3.1, the behavior of 'Mad Max mode' in highway scenarios has become noticeably 'tamed'. Especially in low or no traffic situations, the system will not approach 85 mph and will switch to the right lane several miles before highway exits, much earlier than previous versions. He stated that he will release driving videos to showcase this change, emphasizing it as a positive trade-off between 'safety and habitual driving'.
English review media and several FSD 14.3 owner test reports indicate that the new system under 'Mad Max' configuration has reduced overly aggressive following distances and frequent lane change impulses, while also increasing anticipation in ramp and lane selection, making the system resemble a 'highly skilled but cautious' human driver rather than a 'road barbarian' pursuing extreme speed and challenging rules.
Source: Public Information
ABAB AI Insight
“Mad Max行为变克制”标志着特斯拉在自动驾驶安全与用户体验之间完成了新一轮校准。在早期版本中，“Mad Max”追求速度与机动性，本质上是将“机器的反应能力”作为性能指标，而非“人类驾驶舒适性”作为衡量标准。而在FSD 14.3.1中，系统对车速上限、变道时点与离车流距离的精细调整，显示特斯拉正在把“驾驶的主观体验”作为训练目标的一部分，而不只是“无碰撞”“车道维持”等绝对约束。
从结构上看，这种变化反映出“司机‑算法”之间的权力再分配。在前几版，算法为了尽可能利用其反应能力，倾向于在高速上频繁变道并贴近限速边缘；而在14.3.1，它被“制度化”为“更像老司机”的模式，即提前规划、适度保守、减少突兀动作。这种结构变化不仅能降低驾驶员心理压力，也减少与非自动驾驶车辆间的“意外互动”，使系统更容易在混合交通中形成行为同质性。
长期趋势上，这表明L2‑L3自动驾驶正从“极限性能展示”阶段转向“可解释习惯”阶段。未来用户选择的“速度模式”将不再只是“快慢”问题，而是“系统驾驶风格”与“道路风险‑舒适度权重”的组合。在监管与公众接受度的压力下，越是激进的“机器气质”越会被系统规则压制，而“类人平滑驾驶”将成为自动驾驶能力的“软门槛”。Tesla 14.3.1的“Mad Max驯化”，实质就是为即将到来的“自动驾驶口碑战”铺设的行为准则。