Tesla removes manual speed controls from self-driving software

Tesla is phasing out the manual speed override feature in its Full Self-Driving software, and developers are focusing on AI-driven profiles instead. The company wants the system to drive exactly like you do. Engineers are building per-driver learning directly into the neural networks, so the car remembers your specific corrections when you take the wheel.

The end of manual speed controls

Tesla VP of AI Ashok Elluswamy made the new direction clear on X. He posted that “Max speed control is an anti pattern” and confirmed the old override feature is permanently gone.

He added the team is “working on better learning of user’s implied preferences” for future updates. This marks a clear move away from explicit sliders, and owners notice recent versions have already dropped the right-scroll-wheel speed setting. Drivers just select from modes like Sloth, Chill, Standard, Hurry, and Mad Max to guide the vehicle’s pacing.

How speed profiles operate now

Earlier Autopilot builds let users set a specific offset above the speed limit, but Tesla removed that exact control when transitioning to V14. The system now uses fixed profiles that blend speed choices with lane-change aggressiveness.

Tesla documentation outlines these profiles as the main way to guide the car’s behavior. A profile like Sloth stays close to the limit, and more assertive options drive faster and change lanes more frequently. The scroll wheel now switches between these modes instead of adding a precise speed limit.

This software update aligns with Elon Musk’s recent promises about vehicle memory. Musk recently replied to a complaint about the car exiting the carpool lane.

He stated that “the car will start to remember your specific interventions and match each person’s individual preferences.” He later added that an “upcoming release will remember individual user preferences, so you won’t need to keep correcting your car,” tying the update to fewer manual takeovers.

The CEO teased the car will even learn where you prefer to park at home or work, these features might bundle with other tools later this year.

What implied preferences look like

Engineers have not released technical documents explaining this learning process yet. Third-party analysts believe the car uses reinforcement learning based on your daily habits. If you consistently brake early or choose a specific lane, the neural network treats those actions as feedback. The software gradually adopts your preferred following distance and parking spots over time.

The vehicle feels more like a trained chauffeur than a rigid robot, and observers expect this personalized learning to run alongside the main speed profiles.

AI context over direct input

Elluswamy called the old control an anti-pattern to argue against conflicting AI goals. Giving a user a hard speed limit clashes with an artificial intelligence meant to adapt to changing road context.

The system evaluates speed limits, traffic flow, and navigation data simultaneously to make decisions. The removal of the speed offset caused some complaints from owners. Some drivers received speeding tickets when assertive modes accelerated past posted limits. European regulators closely monitor vehicles that easily exceed speed limits, so the removal of the manual offset aligns well with strict safety rules.

FSD will rely heavily on these broad profiles as the AI learns behind the scenes. Drivers missing the old slider need to adjust to letting the car learn from repeated corrections.

Tesla continues to remind buyers that the software is a Level 2 system requiring constant supervision. You remain fully responsible for the car at all times. The manufacturer clearly envisions a product that adapts to your routine automatically, and we just have to wait for the official release notes to see when these exact memory functions go live.

We’ve launched an official Tesla Owners Online Forum. Join early and claim your "Founding Fuel" badge.

Join Now ⚡