The big picture: given that Tesla FSD can’t drive safely in even light snow, we’re left wondering how Elon Musk can so confidently claim that his “Robotaxis” will be fully autonomous without a safety driver by the end of 2025… which is about ten days away.
Watch FSD Try to Park Itself on the Lawn
Watch this, the Tesla driver has to pump the brakes to stop FSD from trying to accelerate out of a skid, almost hopping the curb into some poor sod’s front yard.
Gave FSD a chance in the snow, we’re far away from full autonomy (14.2.1)
byu/Hardwood_Lump_BBQ inTeslaFSD
My Blow-by-Blow Commentary
TL;DR… yes, this FSD performance blows.
Here’s what we’re seeing:
- Light snow, looks like about an inch of accumulation on the street.
- Road both plowed and salted, per OP. If you don’t drive in snow, that’s optimal and makes controlling the vehicle easy.
- Tesla Model Y in FSD mode biffs the turn, tries to accelerate out of the skid (!?!), and the driver brakes to avoid jumping the curb into some poor sod’s front yard.
So what went wrong here? Well, a lot, by the look of things. Let’s hear the driver’s thoughts first, because they’re very insightful.
Reporting From the Driver’s Seat
Quotes from the driver… or should we call them the FSD Supervisor? God, the future is so confusing!
“The tires are fine, I have them on two other vehicles. The concern here is that the software thought acceleration while fully turned wheels to the left was the corrective action to exit the condition. That is the opposite of what should be done, any driver with any level of experience in weather would do the correct action. FSD needs to learn not everywhere is souther California [SIC]”
Woof. Spot the lie, though.
I’m a native Chicagoan, and if you feel that tell-tale wobble in the tires that signals you’re losing traction on snow, you definitely do not apply the accelerator like FSD visibly did here. The driver (unpaid FSD beta tester?) is right on the money. You decelerate and steer into the skid to maintain control. Every pint-sized Chicagoan with a learner’s permit knows this.
Honestly, I think the driver is also exactly right about the source of the problem. Tesla collects an enormous amount of always-on telematics data from its fleet. It sure looks like they either don’t have nearly enough inclement-weather data, or they’ve programmed the software to squander what data they do have. Neither problem is a quick fix.
“What Tires?!”
Whenever Teslas perform poorly, there’s a race among Tesla fans (or bot farms designed to sound like Tesla fans) to blame external factors.
It’s so common that we now reflexively list things like FSD version and tire make/model whenever we cover the frequent mishaps with Tesla’s self-driving tech, or any cringe-comedy goatrope involving a Cybertruck.
The tires here are the stock tires the Model Y ships with: Pirelli Scorpion Mud + Snow.
The FSD version and hardware are modern: 14.2 on HW4.
The vehicle itself is a 2020 Tesla Model Y.
Don’t believe the Tesla fanboys when they howl that a tire with snow literally in the name isn’t actually sufficient for snow. The driver here reports that the tires perform just fine… when they are manually driving. Here’s the quote.
“I have [the stock tires] on two other vehicles, no issues… I took similar and sharper turns on the same drive manually driving, no issues.”
Something in the FSD software is uniquely poorly suited for driving on snow. Most likely the logic around traction control breaks down in rare-ish but very real weather conditions like light snow or rain.
The Robotaxi Reality Check
There’s no way to remove safety drivers from Robotaxis if this kind of performance is the best Tesla can manage in snow. Austin, Texas does indeed get snow. Not Chicago snow, but enough to matter. So do several of the cities slated to be next up in the Robotaxi rollout.
And if Tesla can’t autonomously drive in light snow, forget about following Waymo into places like New York, where Waymo is already undergoing winter testing.
If Tesla can’t handle an inch of snow in self-driving mode without trying to make like a lawn-dart into the neighbor’s yard, fully autonomous Robotaxis by the end of 2025 aren’t just optimistic, they’re pure fantasy.
