Earlier softwares made you keep your hands on the wheel and it monitored for hand movement regularly to make sure you were paying attention.

Newer software does not require the hand on the wheel. With the older software, I would flick volume up one down one all the time with the steering wheel control because that showed your hand was on the wheel. If you put too much on the wheel, it did cause FSD to disengage which was a pain.

There became a market for a weight people could put on the wheel which would have enough pull to trick the computer into thinking you were touching the wheel, but not enough to disengage the FSD.



Current software uses the camera in the car to monitor your eyes for attentiveness.

GarlandAg2012 said:

---

GAC06 said:

---

You absolutely do not have to keep your hands on the wheel in FSD

---

Yeah I mean it doesn't actually make sense because my understanding is that steering input is one of the ways to take control back from FSD, so if you needed your hands on the wheel to make that happen, I would think people would be deactivating it accidentally all the time. It is odd that the manual says you need your hands on the wheel though. Google says it uses capacitance to check to see if your hand is in contact with the wheel but there are enough people saying it isn't necessary that having your hand off the wheel alone must not deactivate it. Maybe if you do touch it you can look away longer?

At this point I'm just joshing around. I get that FSD is pretty good and in terms of the number of people benefitting from it, it's clearly in the lead. I just think it is less advanced from a pure tech perspective.

---

Old FSD software required you to have your hand on the wheel where it can sense slight resistance. New FSD doesn't and monitors your eyes.

You can disengage FSD with moving the wheel, press the brake, or push the FSD button.

MW03 said:

---

As for the comments about enjoying driving and the like, I love to drive and I have always enjoyed having a sporty car or sedan to zip around in when the roads are open and there are curves to be had. However, I do not enjoy driving in stop and go traffic. One of the main selling points to this thing (aside from the amount of car you get for the money) was the possibility that it could destress my morning and afternoon commutes. The second selling point was that it's a fun, zippy little car that is still fun to drive when the occasion calls for it.

Now, there is something primal about the roar of an engine, the rumble in the tailpipe, that feeling when a turbo engages, downshifting into a curve and accelerating out, etc. Nobody can deny that and this car doesn't have those things. But for pure driving, I am not sure there is more you can get at this price point on the market. Granted, that's largely because of the government credit. However, I was also paying $80 a fill up at the gas station for 93 octane before this thing. I wonder if it still wouldn't make financial sense even without the subsidy.

---

I have owned a BMW 3 series, a X5, MY and now a MX.

From a size standpoint a 3 series = Model 3, X5=My, and X7=MX. At similar price points, there is no comparison of a BMW vs Tesla from a pure speed/torque standpoint.

The MX drives like a mercedes from a comfort/quiet standpoint but can go 0-60 in less than 4 secs.

I drive about 30k miles a year and save about $3k a year in gas/oil changes so a much cheaper car to have.

lancevance said:

---

Medaggie said:

---

I got it for I believe 5K about 5 years ago mostly due to promises and there was no subscription model. Fast forward to buying a MX a month ago, they had a deal where I got one upgrade free which I used to get a 6 row configuration so it costs me an extra 2K.

Subscription is probably the way to go with the current numbers. Buying has negatives including up front costs, and not being able to transfer to another car. So if you total the car, you can't keep it for the next unless they have a rare promotion which I used to transfer a year ago when my MY got totalled.

The problem with subscription is once it becomes a finished product, they will stop the subscription model or increase it tremendously especially if they allow you use it as a robotaxi.

---

Once it works as intended (can work on laptop or watch movie on phone while it drives) they will likely increase it to $500 per month. And people will pay for it.

---

At that point the whole system will just be a subscription service and you won't actually own a car. You'll just subscribe to a service and use your phone/app to schedule rides everywhere. Like UBER but no driver.

You can pay more for a luxury vehicle with better finishes, cooled/heated seats, bigger screens, etc. or less for a basic vehicle that just gets you from A to B. Or the cheapest option which will be more like a bus route (but still go door to door) but will take a little more planning and time since it will go A to C to D to B as there will be multiple passengers. Ride share to work could also be an option to save $$. This whole system would actually make a lot roads obsolete as you need more roads going into town in the morning and more the other way the evening.

The gov. will get involved and screw it up however.

This morning:

• Got a note saying I needed to keep my hands on the wheel and be attentive when using "FSD (supervised)" upon activating the service.
• No warnings, dings, or notes when I didn't have my hands on the wheel.
• Picked up my phone to unlock it and check an email, got a warning that I needed to tug the wheel within 5 seconds or so.

My one evening of FSD seemed to indicate that the car's ability to decide if I was paying attention was worse when I was wearing sunglasses. It still worked most of the time but did tell me a few times that auto detect wasn't working and I needed to bump the wheel. Driving at night with no sunglasses that never happened.

I had sunglasses on this AM, but mine aren't overly dark so maybe it could tell. Who knows. I can tell you that I ordered a little privacy slider thing to switch over when I'm not using the FSD or Autopilot features. I don't think Tesla is going to do anything nefarious, nor will I ever do anything overly interesting; however, I do not like the concept of potentially always being watched, be it by a laptop camera, cellphone, or now a car.

https://electrek.co/2025/03/16/tesla-autopilot-drives-into-wall-camera-vs-lidar-test/

He also completely lied about it being in FSD/Autopilot. Autopilot was not engaged when he hit the wall.

He tried engaging it a few times and the Tesla rejected him (it made a sound), then it allowed him when he was very close, but he was disengaged before he hit the wall.

My guess is the Tesla cameras could tell if you let it go from a long way away, but couldn't tell if you started it really close to the wall. And he gamed it to get the outcome that he (and his sponsors) wanted.

Right before hitting the wall:


As he hits the wall:


Additionally, the water test he did was absolutely not done with FSD or AutoPilot. It does not let you engage it if you aren't in a lane. You absolutely cannot be in autopilot or FSD and drive right now the middle of a double yellow. It is not possible. You can't even engage it.




The video was sponsored by a LIDAR company and the first half of it was about how awesome LIDAR was.

Additionally, one of Mark's big continual sponsors is Rivian:



The title is about "Self Driving" but he says he is using AutoPilot.

The owner of this LIDAR company is Rober's old friend. Austin Russell gave him $4 million for his TeamSEAS project with Mr. Beast.

WE FREAKING DID IT!! Thanks to the 600k+ individual donors from 191 countries (and some help from my buddy Austin Russell for the final push) we hit our goal on removing 30M of trash from the ocean! This is a huge W for the turtles and everyone who helped in any way. Thank you!! pic.twitter.com/CenxFUKibs

— Mark Rober (@MarkRober) January 2, 2022

Not going to surprise me if President Musk files a big lawsuit over this.

Man I just thought it was a funny headline. Pretty misleading on Robers part if what you're saying is accurate.

Regardless of whether this is autopilot or FSD, the behavior exhibited in this video is not what you want or would expect:

Here is the raw footage of my Tesla going through the wall. Not sure why it disengages 17 frames before hitting the wall but my feet weren’t touching the brake or gas. pic.twitter.com/ddmeyqO3ww

— Mark Rober (@MarkRober) March 17, 2025

My 2015 f150 would have performed better, as it has radar with auto brake pre-collision detection. Is the Tesla pre-collision system vision based as well?

GarlandAg2012 said:

---

Also, waymo can do construction zones

---

Anecdotally, my team has had to call Waymo support and then local PD twice in the past 2 months for a Waymo that locked up in a construction zone in their operation area. Waymo support will only allow uniform PD to access vehicles to move them, even if they are blocking the only traffic lanes available.

ezmoney said:

---

Yes it's intriguing but do people not enjoy driving, control and combustible horsepower? Serious question.

---

I enjoy driving, but I much prefer plugging in a destination and letting my Tesla do all the driving. It's amazing!

I'm not saying one should do so, but with a set up like this, you could supposedly, theoretically, hypothetically, use your phone while the car and chauffeurs you around. It would think your eyes are on the road.

He doesn't even get into all the other issues I got into above, like the AP/FSD not being possible when you are over the double yellows.

ezmoney said:

---

Yes it's intriguing but do people not enjoy driving, control and combustible horsepower? Serious question.

---

This. If you want a self-driver, you should take the bus...lol...I kid, but I really love to drive, just can't imagine "riding"...

USA*** said:

---

ezmoney said:

---

Yes it's intriguing but do people not enjoy driving, control and combustible horsepower? Serious question.

---

This. If you want a self-driver, you should take the bus...lol...I kid, but I really love to drive, just can't imagine "riding"...

---

I think it depends on how you are wired. I don't use it outside of rush hour, except as a novelty when I have someone riding with me for the first time. I did have an exceptionally long day recently, and I enjoyed putting it on so I could relax my mind a little bit.

That being said, it's worth at least 5-7 blood pressure points during my commute. I am no longer screaming at cars or getting pissed about switching lanes, getting cut off, etc. I just put the car on and it takes care of all the braking, merging, and other a-hole drivers. I'm sure it'll all show up later somewhere else.

MW03 said:

---

USA*** said:

---

ezmoney said:

---

Yes it's intriguing but do people not enjoy driving, control and combustible horsepower? Serious question.

---

This. If you want a self-driver, you should take the bus...lol...I kid, but I really love to drive, just can't imagine "riding"...

---

I think it depends on how you are wired. I don't use it outside of rush hour, except as a novelty when I have someone riding with me for the first time. I did have an exceptionally long day recently, and I enjoyed putting it on so I could relax my mind a little bit.

That being said, it's worth at least 5-7 blood pressure points during my commute. I am no longer screaming at cars or getting pissed about switching lanes, getting cut off, etc. I just put the car on and it takes care of all the braking, merging, and other a-hole drivers. I'm sure it'll all show up later somewhere else.

---

I guess I could see that, if I still commuted. But for the most part, I still enjoyed driving for most of my commute.

Now, 24 miles through rural Comal Co, through Schertz, to Randolph AFB isn't Houston, or D/FW or Austin or SA, so my commute was never really that terrible.

I guess it's like flying commercial, which I hate. I like to be in personal control of my vehicle. Yep, I'm an "old" and I just wouldn't be able to trust it.

I was about to make a joke about being on the lookout for Wil-E-Coyote ambushes but then I clicked the link to the article and saw they beat me to it. Very realistic, valid, and useful "test"

So I agree that the video is not an accurate test.

But the video aside, aren't there real life examples of Teslas involved in crashes, including fatalities, in autopilot/self driving in situations where lidar likely would have prevented it?

Tesla Cybertruck with HW4 using FSD v13.2.8 did not hit the wall in a new recreation of Mark Rober's "Wile E. Coyote-Style" self-driving test. However, a Model Y with HW3 using FSD 12.5.4.2 was going to hit the wall.

Nice test @alsetcenter https://t.co/1P6tzxbNFW pic.twitter.com/vElR1NP7RU

— Sawyer Merritt (@SawyerMerritt) March 20, 2025

HW4 passed. HW3 failed.

htxag09 said:

---

So I agree that the video is not an accurate test.

But the video aside, aren't there real life examples of Teslas involved in crashes, including fatalities, in autopilot/self driving in situations where lidar likely would have prevented it?

---

There are certainly accidents.

https://www.tesla.com/VehicleSafetyReport

But it seems like fewer accidents with autopilot than typical drivers have. Autopilot < FSD. It is older (but more available) technology.

https://static.nhtsa.gov/odi/inv/2022/INCR-EA22002-14496.pdf

This 11mo ago.

From Aug 2022-Aug 2023 there were 60 FSD crashes and 1 fatal crash. This was the much earlier version of FSD, which is significantly inferior to the current one.

I think that was the crash where the Tesla driver had a .27 BAC, but am not sure.

I think that most agree that in many situations LiDAR is better than visual, but there is also a significant additional cost that most people would not be willing to foot the bill on. Ideally, a car would use visual, LiDAR and radar. But again, cost.

But there are also a lot of situations where LiDAR doesn't work - like any sort of moisture. That's why the Rober video didn't have the car wet - just in front of them. If the LiDAR was wet, it messes up the ability to read the signals because of refraction. Same with fog. Real fog gets your car wet, smoke machine briefly - not so much.

Smarter move would be to ask him to come ride in the Cybercab and struggle session him in to saying it's awesome.

htxag09 said:

---

So I agree that the video is not an accurate test.

But the video aside, aren't there real life examples of Teslas involved in crashes, including fatalities, in autopilot/self driving in situations where lidar likely would have prevented it?

---

Got a car computer with a crash on FSD Beta (v11.4.7, gw 2023.7.30) in the mail

The typical "night, disabled car partially in lane you can see from far away" settings.
FSD active

Please be fully alert when using AP! pic.twitter.com/Ir6j8ehke4

— green (@greentheonly) March 20, 2025

Obviously all self driving systems have flaws, this one seems pretty avoidable though. Cruising in the left lane not paying attention in a narrow highway is a bad idea no matter what driver aids you have.

I'm glad Tesla has applied for a robotaxi permit in California, they will have to publish crash data so consumers can actually compare something meaningful vs the competitors rather than anecdotes and contrived "experiments"

V11.4 is 2 years old. At that time they were still using heuristic code. V12 was a major improvement in performance as was v13. Still not utilizing the full capability of HW4 with the current build. Probability of this happening on current build is substantially lower. The errors I've experienced with FSD lately have been almost entirely related to lane choice due to map data errors.

That's good to know, are the updates not automatic?

Were the driver attention systems not more strict when the product was worse? This is why shipping a beta to millions is a risk.

GarlandAg2012 said:

---

That's good to know, are the updates not automatic?

Were the driver attention systems not more strict when the product was worse? This is why shipping a beta to millions is a risk.

---

That video doesn't say when the crash was. It likely was old.

The attention system has always been pretty strict. It was when it first rolled out and is now. But I posted above that the one at the time of that software was based on feeling your hand on the wheel and that people sold weights that clipped on the wheel to get around that.

The current one does not require the hand on the wheel and is much better (camera on driver from above rear view mirror)

And the software now is WAYYYYY better than back then.

They push the software to you, you click install on your phone. It takes between 25-40 minutes to install and you can't drive when it is installed.

So I agree 100% that any Tesla auto driving system is safer than humans. But, right or wrong, that's not really the baseline. People, as a whole, are ok if there is a crash because of a human "accident" that results in a fatality. Less so if it's because of a design fault, technology, etc.

https://www.wsj.com/business/autos/tesla-autopilot-crash-investigation-997b0129?st=WCEuqx&reflink=article_copyURL_share

This is the article I remember reading. Actually, remember something more in depth but can't find it.

I mean let's be real, the entire reason Tesla says things like you have to have your hands on the steering wheel (even if not physically monitored), is to try to give them an out in lawsuits. Much easier to say, well you abused the system so….

I agree with your point with a combo of the two and this issue being cost. But settling these fatality lawsuits also adds up…..

Video is from a crash in September 2023. Guy that posted it apparently buys computers from totaled Teslas, presumably for parts, and then extracts the dash cam video and driver data for the crash. At the time of that crash that version was the most up to date version.

Hard to tell what she's doing in the video/why she didn't see the car, especially because she's Asian. At the time Tesla required eyes on the road and hands on the wheel, intermittently applying slight turning force to the wheel. I didn't have a Tesla/FSD for that version, but my recollection is that when they went to hands off, eyes on the road the driver monitoring system got more aggressive at requiring eyes on the road.

That update didn't go wide until May 2024 and was the update that removed manually coded rules and was entirely trained based upon driver data. It was a big leap in performance. At the time Tesla was emulating their old computer (released 2017) on their new hardware (released 2023) which has substantially more processing power/higher resolution cameras. V13 was released last fall, which updated the software to more fully utilize the new hardware. I don't push the allowances of the driver monitoring, but I don't know that I could get away with what she's doing in that video with the current version.

Expecting a new update in the next 6 weeks that fully exploits the greater processing power, which should provide another leap in capability once it gets refined.