Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
I would just be happy if they fixed some of the basics issues like staying in the same lane when making a left turn or not going straight if the car is in a left turning lane.

These are basic issues that are still present in v12.
My car jumped out of lane to cut in another car on the right lane (the car was originally in the correct lane) on freeway ramp this morning. I had to pull the car back to the original left lane. I saw the other car braking. This is the most serious bug they need to fix first.

Tesla should fix the first 10 serious bugs in the next 3 months before talking about robotaxi.
 
If you're driving manually, that's fine. But someone might want 1 car length or less (because no one will get in front of me, damnit!), and the NN cannot react fast enough with that close of a gap, and will rear end the car in front if they brake hard. Then the driver will say "BUT BUT BUT, the car was driving, it's not my fault!!!".

Best advice I can give - stop caring about things like follow distance. Just relax and let the car assist with driving. If there is a good sized gap in front of you, and people can change lanes in front of you, just chill out and let it happen. I swear you will still get to your destination, even if it's a minute or two later. The guy that cuts in front of you doesn't matter, really, seriously - doesn't matter.

View attachment 1039731
To quote an old boss of mine, "I don't disagree", though I've been around long enough to know how I'm comfortable driving in various situations thank you. This is another of the oft-quoted "wife-test" fails for me so it's up on my priority list... :cool:
 
This is the type of smooth stopping behavior people here enjoy.

In all these cases eliminating brake use and a smoother stop could be accomplished by either using the accelerator, or just disengaging FSD and coasting to a stop. In ALL of these cases. The accelerator is a nice solution in many cases but you have to know it is possible to stop in time.

Everyday stuff. Routine. Some stops are fine, nearly perfect. Many others are like this. Anyway this I think clearly captures what many people are complaining about. Seems like an easy problem to fix but apparently not.

This is the quintessential FSD 12.3.x stop:

Others:


Yellow light goodness; is figuring out the controlling light the next frontier in AI? L5 next year?

 
Last edited:
I'm not liking this pure robotaxi direction FSD is heading in. Not everyone wants to make their personal car available to any complete stranger. Do I really want someone hailing my car after a night of drinking and barfing in it?
Don't worry. When FSD provides robotaxi capabilities, you'll still be the only one barfing in your car. They aren't going to force owners to include their car(s) in anyone else's robotaxi fleet. There's also no worry that it will happen anytime soon. The August announcement by Elon about robotaxis will undoubtedly be some kind of plan forward, a discussion of the robotaxi vehicle design and stuff like that. The software that drives Teslas around without supervision just doesn't exist yet.

When robotaxis are operating, it may be that ordinary Teslas cannot participate in a robotaxi service. They won't have self-closing doors, they'll have steering controls, and they'll be limited to the current compute and sensor hardware. That hardware is thought to be insufficient for a robotaxi.
 
I'm not liking this pure robotaxi direction FSD is heading in. Not everyone wants to make their personal car available to any complete stranger. Do I really want someone hailing my car after a night of drinking and barfing in it?
Nope, but as a couple with a single car and my daughter's family being totally car-free, car sharing with Robotaxi shunting the vehicle between us would be wonderful especially since we walk to 70% of our weekly errands so she could use the car one day a week for her errands and we'd never miss it. Or sending the car off to park itself after dropping me at a medical appointment door would be great to avoid paying the high price of hospital parking when there is free parking on the street just a kilometer away. It could send itself to for the seasonal tire swap with me including a valet key in the the car for the tire jockey to use once the car got to the garage.

All of those things are big time/energy savers and would help keep our family's carbon footprint very small.

And they were in the back of my mind as a pipe dream when we bought the car and my husband wanted to pay $$$$$ for FSD. I knew it wasn't ready yet, but wondered if it would come before the car ceased to 'live'. It still has 4 years before my usual 'sell and trade up' cycle so maybe it will despite me being an unbeliever for the past 3.5 years.
 
Don't worry. When FSD provides robotaxi capabilities, you'll still be the only one barfing in your car. They aren't going to force owners to include their car(s) in anyone else's robotaxi fleet. There's also no worry that it will happen anytime soon. The August announcement by Elon about robotaxis will undoubtedly be some kind of plan forward, a discussion of the robotaxi vehicle design and stuff like that. The software that drives Teslas around without supervision just doesn't exist yet.

When robotaxis are operating, it may be that ordinary Teslas cannot participate in a robotaxi service. They won't have self-closing doors, they'll have steering controls, and they'll be limited to the current compute and sensor hardware. That hardware is thought to be insufficient for a robotaxi.
I was reacting to the notion going around the community that FSD is being optimized for robotaxi. If they are going to develop a separate vehicle for robotaxi, they should keep the software separate from the L2/L3 cars. Like you said L4/L5 is not coming to our consumer cars anytime soon if ever, so what good does it do to remove user input to the car. Though I think in reality the reason we are losing these option is because of NN not robotaxi anyway.

Also I don't barf in my car, sounds like you are doing some projecting there.
 
This is the type of smooth stopping behavior people here enjoy.

In all these cases eliminating brake use and a smoother stop could be accomplished by either using the accelerator, or just disengaging FSD and coasting to a stop. In ALL of these cases. The accelerator is a nice solution in many cases but you have to know it is possible to stop in time.

Everyday stuff. Routine. Some stops are fine, nearly perfect. Many others are like this. Anyway this I think clearly captures what many people are complaining about. Seems like an easy problem to fix but apparently not.

This is the quintessential FSD 12.3.x stop:

Others:


Yellow light goodness; is figuring out the controlling light the next frontier in AI? L5 next year?

I've been using the accelerator but sometimes I get the needless audible alert that I'm not slowing enough - even though I'm closing the distance with a steady decel and plenty of remaining distance.
 
Watched each one and none of these stops bothered me very much. I suspect if you had someone in the back seat they would never even notice.
Seeing as tons of people here have the same complaint, I doubt this very much. All super noticeable and abnormal as has been confirmed repeatedly here. That is why I posted the videos - makes it super super clear what the issue is.

I've been using the accelerator but sometimes I get the needless audible alert that I'm not slowing enough - even though I'm closing the distance with a steady decel and plenty of remaining distance.
Yeah have to be careful that regen is at a maximum. Usually after avoiding the initial sharp hit using the accelerator, you can just let the car take care of most of the rest. Just have to be conservative. In some cases of very late reaction, it is not possible of course.
 
FSD has long been able to cross double yellow lines to go around stationary obstacles (like delivery trucks).

I saw something today I haven't seen documented anywhere yet: it tried to make a pass around a moving vehicle across yellow lines into oncoming traffic. On a 65 mph highway with double yellows, no less (!).

(Side note: on rural highways in Colorado the speed limit detection and logic for using V12 vs V11 end to end are pretty unreliable. Normally V12 isn't engaged in 65 zones.)

It wasn't successful (didn't accelerate nearly enough for the pass) so I took over. I'm just pretty surprised to even see it attempt that. No interior footage but I can share the dashcam footage if there's interest.
 
Seeing as tons of people here have the same complaint, I doubt this very much. All super noticeable and abnormal as has been confirmed repeatedly here. That is why I posted the videos - makes it super super clear what the issue is.


Yeah have to be careful that regen is at a maximum. Usually after avoiding the initial sharp hit using the accelerator, you can just let the car take care of most of the rest. Just have to be conservative. In some cases of very late reaction, it is not possible of course.
While I’ve seen you post tons on this topic not sure I would qualify that as all? I watched each video and did not see a single one I would be uncomfortable with. I don’t need to come in Hot at a stop. What is the hurry to race to stop any way.
 
I believe the honey bee does not know anything but is acting on instinct.
To paraphrase Arthur C. Clarke: “Any sufficiently advanced instinct is indistinguishable from intelligence.”

One useful definition of intelligence might be: “the ability to solve problems one hasn’t encountered before.” Each driving situation is unique, though many are similar. With enough examples in its training set, FSD ought to be able to extrapolate to solve _nearly_ all driving tasks. Of course, “nearly” carries a lot of weight here. Currently it’s around 99% or so. It will need to reach 99.99999% to become suitable for L4 autonomy.
 
While I’ve seen you post tons on this topic not sure I would qualify that as all? I watched each video and did not see a single one I would be uncomfortable with. I don’t need to come in Hot at a stop. What is the hurry to race to stop any way.
100% agree!!! One of the key shortcomings demonstrated by the videos, as you astutely point out.

That’s part of the problem which I did not highlight. I’d start stopping sooner than FSD and stop stopping later. When driving manually.

Fortunately it is possible to also stop quite fast with regen - the route I take when using FSD, by pressing the accelerator, leading to a smoother, gentler, more relaxing stop than FSD executes. Counterintuitive but the accelerator works marvelously to stop gracefully.

Unfortunately with FSD I do not have the luxury of easing off earlier as I would when coasting to an obvious red light. See the first video for an example of FSD not doing this. Definitely one of the things I miss when not driving manually.

I guess I could post videos of how I would do it but it seems useless to do so if people cannot see the glaring problems demonstrated incontrovertibly by these videos. We all know what regen plus braking going balls to the wall feels like. And we know it rarely happens when driving manually. If I wanted that, I’d be autocrossing!

The hesitantly for 12.3.4 to make turns either from a stop sign (left it right) or a right on red is ridiculous. The view isn’t obstructed, and there is no traffic, it just sits there, inching forward, taking forever to complete the maneuver.
Yep. Broken. One of the original complaints in my first v12 review. Nothing has changed.

Basic functionality broken.
 
Last edited:
  • Informative
Reactions: primedive
Does anyone else in this thread have two Teslas? If so, are you noticing this: I have two 2020 Teslas, an AWD 3 and a MYP. Both have been on identical versions of V12.x, now both on V12.3.4. The Y is much, much better at FSD. I have no idea why, but I very rarely have a safety related intervention with the Y, but with the 3 it's a much more common experience. I have re-calibrated the cameras on the 3, no help. I swear, it's like entirely different versions of FSDS on each.
 
Does anyone else in this thread have two Teslas? If so, are you noticing this: I have two 2020 Teslas, an AWD 3 and a MYP. Both have been on identical versions of V12.x, now both on V12.3.4. The Y is much, much better at FSD. I have no idea why, but I very rarely have a safety related intervention with the Y, but with the 3 it's a much more common experience. I have re-calibrated the cameras on the 3, no help. I swear, it's like entirely different versions of FSDS on each.
I've wondered about that. I have a 2023 MYP with HW4. It doesn't run into curbs or run over lines, and it's generally been pretty good. Almost the only time I disengage is to go faster.
 
Since we are mostly in "holding" for 12.4.x and are slightly OT, so here is my "to ∞8/8 and beyond" thinking.

Conjecture/speculation/HUGE leap in assumption/pulling leprechauns out off my s$$

[assumption]2026 Tesla produces and starts rolling out a L4 robotaxi with fairly wide ODDs and what that means to current Tesla owners:

[conjecture] Even if the robotaxi uses the same hardware and software (FSD V14.x??) we use on our cars we will still stay L2 FSDS. We will just have a competent point to point L2 system but it will be L2 FSDS. This is by far the easiest and most straight forward path for Tesla since they don't have to deal with regulations, liability/insurance issues and general public robotaxi problems/expectations/competition.

So it would (could) be:
  • Tesla sells cars that are L2 FSDS and can take you point to point
  • Tesla makes, owns and operates a fleet of robotaxies
 
  • Like
Reactions: FSDtester#1