Tesla Robotaxi Videos Show Speeding, Driving into Wrong Lane

Watching one of the videos, the car went 38/39 in a 30, and was the only car nearby. It was not keeping pace with other vehicles in its lane. It seemed like a lot of the roads it was driving on were signed 35 or 40 (it was a lot of stroads), but a few were signed 30, which it just flatly ignored.

True, but AFAIK Waymo strictly follows speed limits.

> If the speeding is only going with the flow of traffic, that's not a huge deal imo.

If it's intentionally programmed to violate speed limits, then sure.

If it's intending to follow speed limits but is failing, then that's terrifying, because where is its wrong information coming from, and how catastrophic could the errors be?

"We're at the corner"

Um ... kinda...

Nah, I want a system that obeys the law. I don't need software breaking the law on my behalf. Waymo has proven that it's an exceptional ride experience when you have an automated driver that is cautious and obeys the law, which is more than what 99% of drivers do in the Phoenix metro. Go count how many people obey traffic laws for fun.

And guess who's at fault when you're caught?

Edit: No, crossing double-lines to avoid obstructions isn't breaking the law. People in the Comma Discord bring up these weird edge cases that are outlined in law, too. No, you're never forced to break the law when driving.

You can downvote me all you want. You're wrong. The police would find you wrong, a traffic law court would find you wrong. People are animals and don't want to obey the law. Just obey the law. Drive the speed limit, keep right unless passing or turning left.

I think there are two distinct types of "bad" here, and to me, one is way more interesting than the other.

If we mean, that this is dangerous: Yes, of course. It's an obvious, somewhat dangerous error (and I say "somewhat", only because I assume currently everyone participating is cognizant of the fact, that this is in some kind of testing stage)

But I think the more interesting question is: How quickly can this issue (and others like it) be fixed? If the answer at this point is "we will have to retrain the entire model and just hope for the best" that sounds, like, really bad.

As I write this comment, it's up 8.23% today alone (from ~322 to ~348).

TSLA is a meme stock at this point to which ordinary expectations do not apply. The price is unmoored from what the company does. _By the book_ I agree, it should be shorted, but I wouldn't.

Tesla is going bankrupt or bought by a competitor at some point and Elon is going to prison at some point. I would be shocked if that's not the case. It's all smokes and mirrors, manipulation, deception and it will eventually come back to earth.

However, there is a little saying - 'Markets can remain irrational longer than you can remain solvent" that you seem to want to learn the hard way !

PS. I'm short the stock - but it's a hugely risky bet that I would not recommend to anyone.

I just shorted it as well.. 3 days before you.

And I got roasted. Invest with caution.

Good luck with that. Net profits down 71% YOY? Stock goes up 10% on the news. You think it's going to go down because Robotaxi sucks? Someone will be along shortly (as they always are) to reuse a cliche about irrationality.

> This premature move in releasing the robotaxi is certainly stock pumping.

It is. Inevitably, when the promised rollout ramp up doesn't happen, there will be another pump. This is not his first rodeo. The only thing that can stop another pump is competitors scaling first, not repeated Robotaxi failures (which I fully expect).

A lot of people have shorted TSLA and lost a lot of money - See Jim Chanos. I actually agree the price is going to go down before it goes up, but I wouldnt waste money trying to time it

Long term I think they are fine - I think they will have solved FSD and no one will remember Elon's politics.

And the market seems to love it for some reason. I don't get it. Waymo's been doing the robotaxi thing w/o any employee in the vehicle for several years now. Tesla's launch yesterday is laughable in comparison.

It surely means these things will never operate safely in rain, snow or even bad fog. That seems like a pretty big issue for a taxi operation.

Did they ever have LIDAR? I completely agree that the removal of radar was a cost-cut too far and a decision made with uneducated hubris (his deeply-flawed "humans only need vision so cars only need that too" thought process). If they ever want to actually compete in this space, I expect we'll see one of these (or both) quietly re-added.

They could still add it later, maybe once the solid state ones become more affordable?

Driving over a double yellow is expected and legal in normal driving, such as when making a left turn or going around an obstruction.

In this example it looks like it oscillated between two different routing choices (turning left and going straight), and by the time it decided the correct route was to go straight, it found itself misaligned with the lane it should have been in. Instead of moving all the way back to the right, it kind of “cheats” its way into the upcoming left turn lane. This isn’t something it should do in this situation, but it’s likely emulating human behavior seen in situations like this which appeared in its training data, where people cut across the center line(s) ahead of the turn lane forming when they can see that it is clear.

The thing a lot of people get wrong is that they think the most valuable data for Tesla to collect are the mistakes or interventions. Really what they need the most is a lot of examples of drivers doing good job of handling complex situations. One challenge, though, is separating out those good examples from the less good or bad ones, as human drivers are notoriously bad at, well, driving.

3 billion miles of two dimensional 1080P images is what they've got plus telemetry for speed, acceleration, etc. A lot of the imagery was acquired at night or in foul weather or taken by a camera with a dirty or broken lens.

If it turns out that lidar is necessary, there goes the value of the installed base of Tesla, vehicles. It would be like starting from scratch. Clinging to vision only AV is some variant of the sunk cost fallacy.

If you're thinking, surely they must've thought of all that and have a plan, I'd point to the design of starship and ask, is that ever going to be human rated?

Waymo is working great with far less miles (40 million?). I use it all the time. And know many other friends who use it all the time and love it.

Until they are no longer needed in the vehicle, the safety monitor with his thumb on the kill switch is one per vehicle. Waymo has built out remote monitoring infrastructure, and in car intelligence that enables one remote monitor to cover a large number of vehicles. The vehicles ask for help when they are unable to confidently make a decision about how to proceed. The remote monitors never actually steer a Waymo vehicle. They pick from a set of choices, or they decide to take the vehicle out of operation.

In the short to medium term, safety monitoring will be the limiting factor.

"product launches" - keep in mind this is Tesla's product launch.

They launched Tesla Semi 8 YEARS ago, and the truck is nowhere to be seen. Not even official specs have been released.

This product launch was for influencers only, paying 4.2$ per ride.

I'm not gonna lie - I hate the world where it's possible to become the richest person in the world, while being a snake oil salesman.

You are missing a key factor: Waymo. Tesla has to deliver a product that is at least as good as Waymo. If they can’t do that, why would a person switch to Tesla?

Waymo has a multi-year head start; that’s going to be hard to overcome without a superior experience from Tesla.

As certain as forgotten about in a week’s time.

A fun anecdote - a lot of people may remember Roomba from forever back with their automated little vacuums. Roomba's market share declined significantly because they failed to adopt Lidar technology as quickly as their competitors, instead they depended on the bumper for as long as possible. This put them at a disadvantage in navigation and efficiency as their competitors started using Lidar. Combined with aggressive pricing from rivals, expiry of its patented roller in 2022, a weird insistence to not combine vacuum and mopping into one device, Roomba (or iRobot now) is a just little fish in the sea it made.

I mostly agree with you. I use both FSD on my Tesla & Waymo regularly (LA region), and Waymo just feels way safer in comparison. While FSD has improved significantly the last few iterations, I have seen it do "strange" things often enough that I don't feel safe just sitting in the backseat like I would with a Waymo.

Even if it's hypothetically 99% as good as Waymo at the moment, 99% is not good "enough" when it comes to something as critical as driving.

Today’s “FSD” has its limitations and requires supervision, but your description of it is not anything like my experience even on a HW3 vehicle. In fact, in many years of using Autopilot and various “FSD Beta” and “FSD (Supervised)” versions for several tens of thousands of miles I’ve literally never seen it “slam on the brakes suddenly for shadows” or “veer into the wrong lane”. I’m not a cult member and my next car won’t be a Tesla because I cannot support Musk after the horrible things he has done these last 2-3 years, but “FSD” is phenomenal when used appropriately and with the right expectations about what it is and what it isn’t. And it has improved a ton over the years, too.

The end-to-end solution was a real game changer, and while the previous solution was still useful and impressive in its own right, moving to the new stack was a night and day difference. With V13 finally taking advantage of HW4, and all the work they’ve been doing since then (plus upcoming HW5 introduction), it’s totally within the realm of possibility that they achieve viable L4 autonomy beyond this kind of small scale demo (and I hope some form of L3 maybe on HW4 before long for customer vehicles).

It's probably also got to do with some magic behind the curtains; Waymo hasn't expanded much beyond their initial regions, so they've been static for years in terms of region. They very well may have hand-crafted expected paths, and obviously as the region coverage goes up that kind of hang-crafting doesn't scale (due to changes in the environment, at least) economically. So we can't really say how much work Waymo is putting into each mile. That's true of Tesla also, except that kind of work is totally antithetical to their entire approach from the very beginning, so it would be really surprising to find Tesla getting stuck in that specific local minima, wheras we can almost expect it from Waymo.

As a counter anecdote, I do use FSD with my family in the car, I also have used it on snowy roads, logging roads, and it does quite well. Not unsupervised well, but better than I expected given that I'm running FSD on a nearly 6-year old car. The number of trips around town that have been totaly interventionless has definitely been going up lately, and usually interventions have been because I wanted to be more aggressive, not because the car was making a major error or even being rough.

I have a Model 3 and it drives amazingly well, in Florida of all places. Ive taken it all over back roads in FL/GL with a little bit of AL and MS as well. The issues you described were much more prevalent when I got my car in '23, and I have genuinely been watching them become fewer and further between as time passed. I've driven at least 10000 miles on it in the last two years and I have only had to intervene twice.

I have no motivation to be positive; I own no Tesla stock or position and just like it because its the best car for me currently. I cannot emphasize enough just how different my lived experience has been from how you describe it.

Shift some nouns and this is basically exactly what LLMs for coding are like, except the downside risk is “git revert”

I have had the exact opposite experience. I literally dont drive anymore and I never have phantom breaking problems. Im on HW3

> doesn’t even understand basic concepts like school buses or trains

Yeah, it would be hilarious if it wasn't so horrifying, I remember watching a level crossing be represented as a weird traffic light that would go from red to off to red erratically, with a similarly erratic convoy of trucks representing the train.

Mind you I remember people claiming FSD was "nearly done" because they'd "tackled all the hard problems, and were now in clean up", and how as a result that meant they could let their FSD take itself through a roundabout, not just straightlining it through. Never underestimate the power of denial.

He's essentially admitted as much:

https://www.usnews.com/news/top-news/articles/2025-06-20/exp...

Tesla's AI is trained on actual driving data, so it is behaving in human-like ways, both good and bad.