FSD 14 now rolling out to employees

[REM]

Over a million people dying every year is a staggering figure! That's with human drivers! If FSD could even cut that in half we would save 500,000 lives every frickin' year! That's the equivalent of saving every resident in the town I live in (population under 20,000), and 24 other towns the same size, every single year!

What we currently have is an unacceptable bloodbath of epic proportions. FSD can't get here quickly enough.

This is the most blissfully ignored statistic in the modern world, and when you ask the average person if they are scared to be on the roadways, they don't blink an eye. It's the most dangerous thing we can do on a daily basis, and people just choose to ignore the real facts.

FSD is already saving lives. Once people figure out how cheap robotaxi is, and they can text and ride, things will accelerate rapidly.

People roll their eyes at me when I try to explain the next 5 years to them.

Can't come fast enough.

Sponsored

 

[Jhodgesatmb]

I don't see how anyone who has followed the progress of FSD over the last 5 years say "maybe in 10 years". That makes zero sense. I would say less than a year.

The functional word was “there”. I am assuming that the responder meant “there” to mean they can fall asleep in the back and always arrive at their destination alive and with no accidents. In that context maybe it will be more than a year but surely less than 10.

 

[Jhodgesatmb]

Has there been any information since the original post about employees or anyone else getting V14? Maybe it was just a typical Whole Mars fanboy comment.

 

[dalton108]

Has there been any information since the original post about employees or anyone else getting V14? Maybe it was just a typical Whole Mars fanboy comment.

Could be. But the normal pattern is that we learn that they’re rolling out to employees, and then it’s pretty quiet until there’s a more concrete deadline in terms of when we might get it. That would be the pattern that I’ve seen since 2020.

 

[txtravwill]

I don't see how anyone who has followed the progress of FSD over the last 5 years say "maybe in 10 years". That makes zero sense. I would say less than a year.

Your opinion but I use it a LOT, and it makes a LOT of mistakes on easy stuff even. It makes deadly mistakes daily for me like right turning into a left turn lane, and turning left from a right turn lane even, it tries to do a clover exit at 70mph without slowing enough (where 20mph is recommended), it crosses the yellow lines constantly on curves, etc.

Even if all that gets better, it still can't interpret complicated parking lot signs and road signs. You will never see banish parking on a public Chicago road with like 500 word signs of when you can/can't park, etc. Having to cross in to the other lane, back then back, etc. due to moving trucks, etc. All that is in the city where it's controlled a bit more. Most of my failures occur more rural, but on very well marked streets and things I wouldn't consider edge cases.

I'll redetermine an estimate perhaps in a year ;-) agree.

 

[Gigahorse]

I had a ride in an 8 year old Model X on Hw3 running an older version of FSD. Was incredibly better that my CT. WHY?

Because the CT is the getting less attention than it should from the FSD team.
People that think the CT is great need to sit in another Tesla with FSD and see the light years in difference.

 

[Gigahorse]

Incorrect.

We have a hardware 4 Model 3 Performance. FSD in Cybertruck is good. FSD in Model 3 is very good.

Both are soon to be excellent, safer than the average human.

I think a lot of the disagreement comes between people's version of good and very good, which is understandable.

The frustration I believe is that the CT is "good" but is the most expensive Tesla, and newest vehicle to the lineup. The fact that it seems like it is behind the SXY is disappointing to most who purchased the CT with the anticipation and implication that it would be Teslas new flagship.

That coupled with the "next year the cars will drive themself" fatigue people have been experiencing for a decade.

Don't get me wrong I won't trade CT FSD for "Blue Cruise" but it certainly feels a LONG way away from the CT allowing me to take a nap or pick the kids of from soccer practice. With the other Tesla models being a ways off, but closer.

 

[PungoteagueDave]

I don't see how anyone who has followed the progress of FSD over the last 5 years say "maybe in 10 years". That makes zero sense. I would say less than a year.

…you are channeling Elon Musk at your own risk…

I have used every iteration of AP and FSD since 2015, passed the FSD test on the first day it rolled out in 2020, and have owned multiple Teslas, currently have two.

I’m fairly confident in saying that neither our cars nor official company robotaxis will ever drive unattended. I use absolutes like “ever”, “never”, “always” rarely and under advisement.

I sincerely hope I’m wrong, find it almost impossible to bet against a guy who lands booster rockets and connects human nerves to machines, but the hardware we have simply isn’t up to it. Just the other day our Juniper MY crapped out in heavy rain with the red flashing screen and “Take Over Immediately” message. This occurred on the Chesapeake Bay Bridge center span during rush hour due to a sudden heavy downpour. The humans slowed down and kept going when the camera-based system was unable to see. If it was a robotaxi it would have stopped in a spot with no shoulder, well over a mile from either bridge end, and would have backed up traffic for 10 miles or more. It simply cannot handle a normal driving situation that occurs with regularity.

Elon promised that our vehicles would drive unattended, and stated specifically that the autopilot hardware two generations back contained everything needed for Teslas to join the robotaxi fleet when owners wanted to make money with their cars. He said that FSD’s value was in its ability to make enough from taxi service to make the owners’ monthly payments. He rhapsodized about being able to put our kids in the car unattended and have them driven to school. None of that is going to happen, either with the cars he said were ready in 2018, the cars we buy seven years later, nor any car that Tesla sells in the next few years. It is simply impossible.

There’s a reason that Tesla’s engineering and FSD software development team is jumping ship. Most are under NDAs, so cannot say much, but it is clear they were handed an impossible goal that they knew then, and ultimately confirmed for sure isn’t possible. There are lots of signals that the wheels are about to fall off the myth machine. Just last month Tesla reassigned several engineers from the Dojo AI project to FSD. Can you imagine having to join a project that is supposedly almost ready to roll, after robotaxi beta began, and getting into the flow with such a Herculean task, with much of the institutional knowledge gone?

Again, I hope this is wrong, and time will tell, but time is now really short - Tesla must put up or shut up. If this is correct, and the truth becomes clear, the accrued financial liability will be massive. I give it six months at the outside.

 

[dalton108]

I think a lot of the disagreement comes between people's version of good and very good, which is understandable.

The frustration I believe is that the CT is "good" but is the most expensive Tesla, and newest vehicle to the lineup. The fact that it seems like it is behind the SXY is disappointing to most who purchased the CT with the anticipation and implication that it would be Teslas new flagship.

That coupled with the "next year the cars will drive themself" fatigue people have been experiencing for a decade.

Don't get me wrong I won't trade CT FSD for "Blue Cruise" but it certainly feels a LONG way away from the CT allowing me to take a nap or pick the kids of from soccer practice. With the other Tesla models being a ways off, but closer.

Correct. The “flagship” positioning is really problematic, IMO.

I was viewing a thread on Reddit today and someone was talking about how impressed they were with their 2026 MX because it had matrix headlights. People were quick to remind them that every S & X since the 2023 refresh has had matrix headlights installed and they were enabled by the 2025.14 firmware update.

That’s before the Cybertruck came out and while THAT is disappointing, what those people don’t know and I (and a few other others) do is that the 2021[!] M3P was the first Tesla in the US to have matrix headlights I got mine in Dec. 2020! It also got adaptive headlights on the 2025.14. Something we don’t, and will never have on our CTs.

That the “flagship” vehicle doesn’t have tech that was on a car I bought FIVE YEARS AGO— from the same manufacturer, makes no fucking sense!!!

So, it chaps my hide a little bit when people talk about the way that Tesla rolls out new features and how big of a benefit is because they don’t wait for “model years” when they do stuff like this. Sure, if they actually did it consistently.

It’s just baffling. There’s absolutely no excuse or reason that they couldn’t have in the last five years figured out how to put matrix headlights in the lights that they decided to design for the CT. For them to say we just didn’t have the room is an unacceptable excuse. Making them what 5% bigger and giving them the latest tech should have been a trivial matter. What the hell?!

I’ll put it different, when I took possession my ‘21 MP3 in December 2020 it had the most advanced tech of any Tesla (by far). Surely, the CT should’ve been in the same position as that car on launch day. It wasn’t!

/end rant

 

[PungoteagueDave]

Robotaxi and actual (Unattended) FSD are most likely destined to go the path of Google Glass. Great idea. RIP Google Glass.

 

[CyberTexas]

Why not wait till V14 comes out before we make conclusion about the future? FSD has advance so fast lately that we can't use past history to predict the future. Tesla AI training has advance leaps and bounds.

Just relax and enjoy CT.

 

[PungoteagueDave]

Why not wait till V14 comes out before we make conclusion about the future? FSD has advance so fast lately that we can't use past history to predict the future. Tesla AI training has advance leaps and bounds.

Just relax and enjoy CT.

Because it CANNOT SEE in a thunderstorm and no amount of software development can fix that. It cannot read complicated multi-row parking lot signs, and no amount of software development can fix that. It cannot deal with gated parking lots, and no amount of software development can fix that. It cannot see or deal with construction flag crews, and no amount of software development can fix that. And on and on…

 

[dalton108]

Why not wait till V14 comes out before we make conclusion about the future?

The answer is: human nature. What do you think these forums are for?

 

[HaulingAss]

Your opinion but I use it a LOT, and it makes a LOT of mistakes on easy stuff even.

Two questions:

1) Have you used FSD on a hardware 4 vehicle that is NOT a Cybertruck?
2) Have you been actively following the progress of FSD over a 4-5 year period?

One additional point:

FSD is quite fault tolerant of camera hardware/wheel alignment variations from one vehicle to the next. It adapts quite a bit, and this means that some cars drive better than others. As it becomes autonomous, I expect Tesla to make the tolerances in tire and camera alignment stricter, so vehicles have more consistent performance from car to car.

 

[Jhodgesatmb]

Because it CANNOT SEE in a thunderstorm and no amount of software development can fix that. It cannot read complicated multi-row parking lot signs, and no amount of software development can fix that. It cannot deal with gated parking lots, and no amount of software development can fix that. It cannot see or deal with construction flag crews, and no amount of software development can fix that. And on and on…

You are dead wrong about weather conditions and signs, which for me calls into question all of your reasoning. For weather it has ample telemetry and information to determine what the local/immediate conditions are. Whether it uses that information while driving is another matter. As for signs, that is just an issue of training. Tesla trained their system to recognize and respond to speed limit signs, so that could just as easily train their system to identify and respond to any sign. I have long wished (and articulated to Tesla) that FSD would recognize and respond to warning signs such as 'reduce speed ahead' or 'windy road next 3 mies'. It isn't that Tesla can't do it but they believe that they are doing a better job from first principles. I happen to believe that they are wrong in that way but this isn't about what I believe is the correct approach but what I believe Tesla can do. It isn't software development friend, it is training, and they 'can' do it. You may be right about gated parking lots. After all, if that requires a card key or some such the best they could do is stop and wait for you to fish out the key and then wait for the gate to open. Same for paid lots where you have to take a card. They could train FSD to wait at the card dispenser until the bar goes up. But why do you think they couldn't at least train the system to do what I conjecture?

Sponsored