Tesla is making a permanent change to how Full Self-Driving handles speed. The company's AI lead has confirmed that fixed maximum speed controls are being dropped entirely — replaced by FSD speed profiles that learn how you like to drive.
What Tesla's AI Lead Actually Said
Tesla's head of AI, Ashok Elluswamy, responded to a post by @DavidMoss on X, calling fixed maximum speed controls an "anti-pattern."
His team, he said, is focused on better learning users' implied preferences — rather than asking drivers to manually input a precise speed limit.
The reaction from the owner community was mixed. Some agreed with the direction. Others pushed back hard, arguing that manually setting a max speed is their explicit preference. Several users criticized the "anti-pattern" framing as a way to avoid genuine product dialogue rather than a well-reasoned design decision.
From Speed Caps to Speed Profiles
Tesla's engineering team is reworking the core logic of how FSD manages vehicle speed.
Instead of a fixed speed ceiling, the system now uses neural networks to factor in user preferences and select a more appropriate driving speed autonomously.
Recent FSD versions replaced the manual speed offset with a set of speed profiles, including the newer SLOTH mode introduced with FSD v14 Lite:
Profile |
Behavior (per user reports) |
|---|---|
Sloth |
Roughly at the posted speed limit |
Chill |
~2 mph over the limit |
Standard |
~5 mph over in most conditions; up to 72 mph on some roads |
Hurry |
~8–10 mph over the limit |
Mad Max |
Largely ignores posted limits |
What Owners Are Actually Experiencing
Real-world feedback from owner forums tells a more complicated story.
Several users find the profiles too coarse. One common complaint: Standard at 67 mph feels too slow, but Hurry at 80 mph feels too fast — with no middle ground available.
Other reported issues include:
Switching between profiles frequently is tedious, especially on variable commute routes
Standard mode sometimes fails to pass slower vehicles proactively
Chill and Sloth can't compensate when the underlying map speed data is wrong
No effective feedback channel exists to tell FSD it's going too fast or changing lanes too often — taking over the wheel is the only signal, and it's a blunt one
Why Tesla Is Making This Change
Two forces are driving this shift: data quality and regulatory pressure.
On the data side, public map and GPS speed limit data is not always accurate or up to date. In the U.S., posted limits often don't reflect actual traffic flow. On a highway posted at 55 mph, traffic may be moving at 75 mph. If FSD follows an inaccurate map limit and falls behind traffic, drivers are forced to intervene — defeating the purpose of the system.
On the regulatory side, Sweden recently urged the EU to reject FSD approval, citing the manual speed offset feature as a speeding risk. France raised similar objections. These pressures gave Tesla additional motivation to remove the manual speed control entirely.
What Comes Next: Learning From How You Drive
In future FSD versions, Tesla plans to replace hard-coded speed inputs with a system that learns your preferred speed from how you actually drive.
Elluswamy confirmed that engineers are building a system to learn implied driver preferences directly from human driving behavior. This aligns with what CEO Elon Musk has previously described: FSD will observe driver habits and interventions.
For example, if a driver taps the accelerator because FSD is going too slowly, the system will log that preference and match that speed on the same road in the future.
Not everyone is convinced. Key concerns from the community include:
Map data errors remain a fundamental problem that learned preferences can't fix
For Tesla's Robotaxi fleet, allowing driver interventions may interfere with the AI's ability to develop optimal autonomous strategies
Without a direct feedback channel, the learning signal is still too indirect to be reliable
The Bigger Picture: FSD's Next Generation
No firm date has been announced for the smarter speed-learning system. But neural network updates are expected to arrive alongside major software version milestones.
Tesla's Robotaxi vehicles are already running FSD v15 — the next-generation architecture with 10x more parameters than the current version. That scale of compute could provide the foundation needed for truly personalized speed modeling.
Meanwhile, Tesla has released updated safety data, with the company claiming FSD is now 5.2x safer than human drivers. Since launch, the system has logged 65 million kilometers in Europe. FSD's global subscriber base has grown from 1.48 million in Q2 2026 to more than 1.5 million, highlighting the continued adoption of Tesla's driver-assistance technology.
The speed cap removal is one of the more visible signs that Tesla is shifting FSD from a driver-assisted tool toward a fully autonomous system — one that observes, learns, and adapts rather than waiting to be told what to do.