Tesla is continuing to reshape how Full Self-Driving manages vehicle speed, with AI chief Ashok Elluswamy confirming that fixed speed caps aren’t coming back to future software builds. Instead of letting drivers manually dial in a strict maximum speed, engineers are training neural networks to account for user preferences and better select ideal driving speeds on their own.
When prominent FSD tester @DavidMoss said on X that he hopes “Max Speed control never comes back to Tesla FSD,” Elluswamy responded directly, writing:
“Max speed control is an anti pattern. We are working on better learning of user’s implied preferences.”
Max speed control is an anti pattern.
— Ashok Elluswamy (@aelluswamy) August 3, 2026
We are working on better learning of user’s implied preferences.
The Problem With Fixed Speed Caps and Mapping Data
Leaving speed selection entirely to software comes with real-world challenges. Public map data and GPS speed limits are not always accurate or up to date. Furthermore, posted speed limits in North America rarely match actual traffic flow, especially in the U.S. It’s common for traffic on a 55 mph stretch of highway to move at 75 mph. When FSD strictly adheres to an inaccurate map speed limit instead of matching surrounding traffic flow, vehicles end up traveling slower than surrounding cars, frustrating drivers and prompting them to set a manual override.
Tesla, however, has been moving away from the fixed Max Speed setting in recent FSD updates, instead switching over to Speed Profiles that let drivers pick general personalities like Sloth, Chill, Standard, Hurry, and Mad Max instead. These profiles seek to match vehicle speed and assertiveness to how the user wants FSD to drive.
Manual speed settings have also drawn regulatory friction. Sweden recently urged the EU to reject FSD over speeding concerns, pointing directly to the manual speed offset toggle that allows vehicles to drive a margin above the posted speed limit. France has also taken issue with the same feature.
Learning Implied Driver Preferences
Rather than relying on hardcoded speed inputs, future builds will learn how fast you like to drive based on how you handle the car. As Elluswamy noted, engineers are building systems that learn implied driver preferences directly from human behavior.
This aligns with earlier teasers from Elon Musk confirming that FSD will remember driver habits and interventions. When a driver taps the accelerator pedal because FSD is moving too slowly, the system will log that preference, learning to match that pace on future trips along the same road.
While no firm release date has been given, these neural network updates could arrive alongside major software milestones. Tesla’s Robotaxi vehicles are already running early builds of the next major architectural FSD upgrade, version 15, which will feature 10x as many parameters as current builds and could power these personalized speed models.
