Fifty-six Miles
Disclosure: I own Tesla stock.
Of the last 9,175 miles I've tracked in my car, I've driven 56 of them manually. Tesla's Full Self-Driving software handled the other 9,119. Recently, I went 220 miles across several trips to work, home, and various stops without touching the steering wheel, pedals, or shifter.
I was in the driver's seat, paying attention. I always am. The car did the steering, braking, accelerating, and navigating. I watched the road and remained ready to take over. Tesla calls the system Full Self-Driving (Supervised), and that last word describes work I still have to do. It is not an autonomous vehicle.1
It is, however, a car that routinely takes me where I'm going while I leave the controls alone. Through complex turns. Around rotaries (which are roundabouts if you live somewhere that doesn't call them rotaries.) To work and back, with other stops along the way. This is a thing I own and use in the course of an ordinary week.
I did not expect to get here.
In 2018, I bought a Model 3. It was my first new car, and I loved it. I had Enhanced Autopilot, Tesla's driver assistance package at the time. There was also an option to pay for Full Self-Driving, which was an opportunity to purchase a promise. I passed.
Then, during a brief sale, the price came down far enough that I bought it. My investment thesis was that Tesla would never pull it off, there would inevitably be a class action, and I'd get my money back. A sophisticated approach to buying software.
When I got access to the FSD beta in 2022, I felt pretty good about the first part of that prediction. The software was terrible. It tried to drive over rotaries. There were moments when its decisions felt actively dangerous. I had bought a promise, waited years, and received something that gave me fresh reasons to doubt the promise.
Version 11 didn't change that feeling much. Version 12 felt safer, but I was still uncomfortable. It would drive faster than I wanted. Being confident that a car probably won't do something disastrous is a long way from wanting it to drive you somewhere.
Versions 13 and 14 have been a different experience. I also replaced my original Model 3 with a new one in 2025, so this hasn't all happened in the same car. During that purchase, I was able to transfer the FSD software I'd already paid for. The car I use now, running the newer software, handles almost all the miles I travel.
The surprise has outlasted the novelty. I expect it to make a complex turn, and it makes the turn. Occasionally, I still find that remarkable.
I'm more willing to go places now. More often, and farther away. Driving in familiar or unfamiliar places no longer leaves me feeling tired or tense. I still pay attention throughout the trip, but watching the road while the car handles the controls feels different from doing all of it myself. That difference is large enough to affect whether I want to make the trip at all. I hadn't expected that part. But a long drive is less of a reason to stay home now.
There are limits I encounter often enough to plan around. In active school zones, I take over. The system's response to the reduced speed limit has been inconsistent in my experience, and I don't leave that decision to it. That is a consequential failure in a place where getting the speed right matters especially.
Parking lots produce a different collection of problems. The car sometimes chooses strange spaces, including accessible spaces it shouldn't use. Sometimes it struggles to find its way out. It also doesn't reliably avoid potholes. A car that can negotiate a complex turn can still make a decision in a parking lot that seems baffling from the driver's seat.
These failures have different stakes, but all of them shape how I use the software. I don't have to decide whether I trust it in the abstract. I have to decide whether to let it continue through the situation in front of me. In some situations, I already know my answer before we arrive.
The mileage figure doesn't capture all that. It tells you how much distance the software handled, not how often I needed to correct it, how serious any particular mistake was, or whether it is safer than another way of driving. A short intervention can matter more than hundreds of uneventful miles. Those 56 manual miles don't measure the importance of remaining ready to drive.
They do tell you how thoroughly this has entered my routine. I bought software I expected to fail. Years later, I reach for it almost every time I go somewhere, with a growing familiarity with both what it can handle and where I want the controls back.
That history makes me optimistic about the problems that remain. I've watched behavior I once thought might be beyond the system become ordinary. I expect more improvement through software updates. I also remember being certain in the other direction, which is a useful reason to hold this prediction a little more loosely. Earlier progress gives me a reason for optimism; it doesn't supply a completion date.
For someone buying access now, Tesla lists FSD at $99 a month in the United States. That's a meaningful recurring expense, on top of owning the car. The company also offers demo drives where you can experience the system yourself.2 If you're curious, I think it's worth trying. A demonstration won't establish long-term reliability, but it can make the capability less abstract. You can sit in the driver's seat, watch what the car does, and notice how that feels.
I still have to watch the road. I still take over. And when I think about going somewhere farther away, the drive feels easier than it used to.
Sources
1. Full Self-Driving (Supervised) — Tesla Support ↗. Tesla requires active driver supervision and states that the system does not make the vehicle autonomous. ↩
2. Full Self-Driving (Supervised) — Tesla Support ↗. U.S. monthly subscription price and demo-drive availability, checked October 14, 2026. Price and availability may change. ↩