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 usually keep FSD off and prefer enhanced autopilot.

So, how do you know the following is true ?

I own FSD. My opinion is FSD drives like an erratic teenager first time driver.

Fact is - people who leave it on understand how FSD "works" and when to use it and when not to. Those who don't just think its junk and doesn't work. It could also be a chicken & egg situation. People try it and invariably get scared by the way it drives and keep it off.

I think its much more capable than a teenage first time driver - though the kind of mistakes FSD makes are different from what a new driver would make.
 
So, how do you know the following is true ?



Fact is - people who leave it on understand how FSD "works" and when to use it and when not to. Those who don't just think its junk and doesn't work. It could also be a chicken & egg situation. People try it and invariably get scared by the way it drives and keep it off.

I think its much more capable than a teenage first time driver - though the kind of mistakes FSD makes are different from what a new driver would make.
I agree. If you treat FSD as really good TACC it works well. Once you learn its limitations, you can anticipate and take over during its limitations.

This strategy tends to work really well on my normal routes since FSD tends to screw up in same situations/roads/intersections.
 
So, how do you know the following is true ?



Fact is - people who leave it on understand how FSD "works" and when to use it and when not to. Those who don't just think its junk and doesn't work. It could also be a chicken & egg situation. People try it and invariably get scared by the way it drives and keep it off.

I think its much more capable than a teenage first time driver - though the kind of mistakes FSD makes are different from what a new driver would make.
Yeah, a teenager wouldn’t pull into a parking lane and wait for the parked cars to move, and a teen could be told to read the overhead signs and lane arrows so you don’t make illegal left turns. FSD junk will just keep doing these things over and over and over again. FSD is more like a 7 year old behind the wheel. I want my money back!!
 
  • Like
Reactions: beachmiles
Yeah, a teenager wouldn’t pull into a parking lane and wait for the parked cars to move, and a teen could be told to read the overhead signs and lane arrows so you don’t make illegal left turns. FSD junk will just keep doing these things over and over and over again. FSD is more like a 7 year old behind the wheel. I want my money back!!
But I'd not trust a teen to make safe lane changes, not run stop signs / red lights, not overspeed, keep proper distance from the car in front, stop promptly for VRU.

If you think FSD is "junk" - it will remain junk and unusable for you. If you think of it as L2+/ADAS+, you can learn how to use it and actually get some benefit out of it.
 
We purchased EAP and are on the free trial of FSD Beta until January. We're on 2023.27.7, and the loaner I picked up today while our car is in for minor warranty service is a USS-equipped Model 3 running 2023.32.9.

In our car on the way to the Tesla SC roughly 33 miles from our house I tried FSD for the first time today. Doing so mostly confirmed for me that our decision to go with EAP vs. FSD was probably a good one. It also confirmed for me that as FSD stands today, my spouse will never use it, and that when she or anyone else are in the car with me I'll never use it anywhere except on controlled access highways.

By and large, FSD and EAP do a fine job on limited access highways where EAP is willing to do it's thing. One caveat: in my trial of FSD today, just as with EAP, it occasionally makes poor decisions in merge scenarios. One, with a large truck climbing the entrance ramp, was handled beautifully. The car smoothly moved over and made room, so nice. However, in a different circumstance where the car could not move over and another vehicle was entering via the onramp, it failed to make any allowances when it should have slowed slightly to allow the car to smoothly merge in front while next car could merge behind us. In one other highway scenario while driving through a construction zone, the car neither tried to move left nor slow down as it approached and drove by construction and police vehicles, all of which had lights flashing. In many states, the law says move over or slow down, but FSD did neither.

On local suburban streets, FSD generally made terrible decisions in the following scenarios, all in full daylight, with excellent weather conditions.
  • Struggled a bit with our neighborhood’s only traffic circle. No cars in the circle. Stopped before entering, and then after entering failed to curve properly in a way that would keep it neatly in the traffic lane. On the way back, the loaner vehicle with a newer software version handled this empty traffic circle better, so that's nice.
  • Couldn’t pick one of two left turn lanes from one street to another when approaching a light. Both streets are divided, and as the car approaches the light, two left turn lanes are added on the left side. The car pulled toward the two left lanes and then randomly swerved back and forth between the lanes, never really committing to either one. So, I seized control and picked one. Fortunately, no cars behind me to get confused or report our car to the cops for such bad driving. White lines were fresh and clear - these are new streets, and for now the painted lines are quite visible. Weather conditions were perfect.
  • As we approached a 3-way intersection controlled by a light in the loaner car, it properly pulled into the left turn lane and prepared to make the left turn, but then right at the end, it suddenly did an early and dramatic left yank on the wheel for the turn, which could have put us into oncoming traffic or head-on into cars waiting patiently at the light so they could get their turn to enter the road we were on.
  • As we approached an unprotected right turn, the loaner car "California stopped" and tried to pull into oncoming traffic when it should have come to a full and complete stop, and waited for traffic to clear before turning right.
  • As we approached an unprotected left turn from an added left turn lane from a 2-lane road onto a divided street with four lanes and a median, the car suddenly pulled hard left and tried to drive right into the lane of cars closest to the median at the stop sign instead of staying to the right of the median. Very very not cool, and pretty darn scary.
So, yeah, technically, FSD tried to crash the car at least 2-3 times on this drive to and from the SC. And no, we'll not pay for FSD any time soon, as EAP mostly performs acceptably on roads where it's available for use and for now FSD seems to be quite a risk on local streets and roads.
 
Last edited:
We purchased EAP and are on the free trial of FSD Beta until January. We're on 2023.27.7, and the loaner I picked up today while our car is in for minor warranty service is a USS-equipped Model 3 running 2023.32.9.

In our car on the way to the Tesla SC roughly 33 miles from our house I tried FSD for the first time today. Doing so mostly confirmed for me that our decision to go with EAP vs. FSD was probably a good one. It also confirmed for me that as FSD stands today, my spouse will never use it, and that when she or anyone else are in the car with me I'll never use it anywhere except on controlled access highways.

By and large, FSD and EAP do a fine job on limited access highways where EAP is willing to do it's thing. One caveat: in my trial of FSD today, just as with EAP, it occasionally makes poor decisions in merge scenarios. One, with a large truck climbing the entrance ramp, was handled beautifully. The car smoothly moved over and made room, so nice. However, in a different circumstance where the car could not move over and another vehicle was entering via the onramp, it failed to make any allowances when it should have slowed slightly to allow the car to smoothly merge in front while next car could merge behind us. In one other highway scenario while driving through a construction zone, the car neither tried to move left nor slow down as it approached and drove by construction and police vehicles, all of which had lights flashing. In many states, the law says move over or slow down, but FSD did neither.

On local suburban streets, FSD generally made terrible decisions in the following scenarios, all in full daylight, with excellent weather conditions.
  • Struggled a bit with our neighborhood’s only traffic circle. No cars in the circle. Stopped before entering, and then after entering failed to curve properly in a way that would keep it neatly in the traffic lane. On the way back, the loaner vehicle with a newer software version handled this empty traffic circle better, so that's nice.
  • Couldn’t pick one of two left turn lanes from one street to another when approaching a light. Both streets are divided, and as the car approaches the light, two left turn lanes are added on the left side. The car pulled toward the two left lanes and then randomly swerved back and forth between the lanes, never really committing to either one. So, I seized control and picked one. Fortunately, no cars behind me to get confused or report our car to the cops for such bad driving. White lines were fresh and clear - these are new streets, and for now the painted lines are quite visible. Weather conditions were perfect.
  • As we approached a 3-way intersection controlled by a light in the loaner car, it properly pulled into the left turn lane and prepared to make the left turn, but then right at the end, it suddenly did an early and dramatic left yank on the wheel for the turn, which could have put us into oncoming traffic or head-on into cars waiting patiently at the light so they could get their turn to enter the road we were on.
  • As we approached an unprotected right turn, the loaner car "California stopped" and tried to pull into oncoming traffic when it should have come to a full and complete stop, and waited for traffic to clear before turning right.
  • As we approached an unprotected left turn from an added left turn lane from a 2-lane road onto a divided street with four lanes and a median, the car suddenly pulled hard left and tried to drive right into the lane of cars closest to the median at the stop sign instead of staying to the right of the median. Very very not cool, and pretty darn scary.
So, yeah, technically, FSD tried to crash the car at least 2-3 times on this drive to and from the SC. And no, we'll not pay for FSD any time soon, as EAP mostly performs acceptably on roads where it's available for use and for now FSD seems to be quite a risk on local streets and roads.
FWIW, I traded in a 2018 M3 running FSD-b 11.4.7.2 for a new M3 running the load that you've got, which turned out to be (I think, it's been a few weeks now) 11.3.7 or something.

I've been beta testing FSD-b with the crowd since March of 2022. I had forgotten just how bad the 11.3.x releases were; better, mind you, than the 10.x releases (by far), but, man: Swinging wide right on left turns, jerking the steering wheel generally on turns, and doing Other Fun Stuff that would scare the gee-willikers out of passengers and other drivers on the road.

After a bit, the car got updated to 2023.32.7, then 2023.32.9, both of which have FSD-b 11.4.4. Mind you, the people who haven't done trade-ins or what-all are still on a 2022.x.x load and are running 11.4.7.x; they complain about that, too (why not) but nobody claims that the 11.4.7.x branch is better than the 11.3.anything branches.

I will report that the 11.4.4 FSD load still has some foibles, but it doesn't scare people and handles turns quite a bit better. Still swings a bit to the right on left turns and sometimes jerks a bit on turns in general; but got back this weekend from a stint from New Jersey to Boston and back with no particular white hairs.
 
I should have said that our 2023 Model YLR (HW4) running 2023.27.7 is running FSD Beta 11.4.7.3, and the loaner Model 3 running 2023.32.9 is running FSD Beta 11.4.4. I don't know how these version numbers manage to juxtapose themselves (e.g. Version 2023.27.7 in our car sounds older than Version 2023.32.9 in the loaner, while FSD 11.4.7.3 in our car sounds newer than 11.4.4 in the loaner), but there it is. Makes it really challenging to compare and reach empirical conclusions about which FSD version is newer and hopefully better. 🤷‍♂️
 
I should have said that our 2023 Model YLR (HW4) running 2023.27.7 is running FSD Beta 11.4.7.3, and the loaner Model 3 running 2023.32.9 is running FSD Beta 11.4.4. I don't know how these version numbers manage to juxtapose themselves (e.g. Version 2023.27.7 in our car sounds older than Version 2023.32.9 in the loaner, while FSD 11.4.7.3 in our car sounds newer than 11.4.4 in the loaner), but there it is. Makes it really challenging to compare and reach empirical conclusions about which FSD version is newer and hopefully better. 🤷‍♂️
All right. As far as the version numbers go: Near as anyone can tell, roughly around Xmas each year Tesla usually manages to get everybody on the same overall version number (like, 2022.12.x or something). Then, incidental fixes and improvements that don't involve FSD go out to everybody not doing Beta Testing; hence, that crowd ends up with higher numbers, like 2023.ww.version1.subversion2 and so on.

The crowd that is doing beta testing gets stuck on an older major release which is slowly incremented; hence, the "2023.27.7". ".6" was probably FSD 11.4.7.2; one can get a better idea by going over to the teslafi.com firmware tracker, which has snapshots of all of this.

On top of that, this year, Tesla is delivering all cars with some version or other of FSD; that's new. And the major releases like 2023.32.9 are going to everybody, and all of those have a software-unlockable version of FSD, currently 11.4.4. So, yeah, confusing.

FWIW, plenty of people over at the 11.x discussion have been complaining that the 11.4.7.3 has lots of "regressions", i.e., poor behavior. But there's no backing up. 11.4.4 was fairly stable and, I'm guessing, is probably why Tesla's handing that out to all.

There's been some guidance by Tesla that they want to deliver a Very Different version of FSD that's end-to-end Neural Networks. The new NN is supposed to replace several hundred thousand lines of C++ code, provide a better driving experience, and hopefully will quiet down some of the nay-sayers. Since this release, nominally 12.x.x, is on Elon time, it may come out 4th Q of this year or 1st Q of next year. Seeing as it's the usual research project, and research is termed, "The process of running up alleys to see if they're blind.", those rough dates are malleable.
 
We purchased EAP and are on the free trial of FSD Beta until January. We're on 2023.27.7, and the loaner I picked up today while our car is in for minor warranty service is a USS-equipped Model 3 running 2023.32.9.

In our car on the way to the Tesla SC roughly 33 miles from our house I tried FSD for the first time today. Doing so mostly confirmed for me that our decision to go with EAP vs. FSD was probably a good one. It also confirmed for me that as FSD stands today, my spouse will never use it, and that when she or anyone else are in the car with me I'll never use it anywhere except on controlled access highways.

By and large, FSD and EAP do a fine job on limited access highways where EAP is willing to do it's thing. One caveat: in my trial of FSD today, just as with EAP, it occasionally makes poor decisions in merge scenarios. One, with a large truck climbing the entrance ramp, was handled beautifully. The car smoothly moved over and made room, so nice. However, in a different circumstance where the car could not move over and another vehicle was entering via the onramp, it failed to make any allowances when it should have slowed slightly to allow the car to smoothly merge in front while next car could merge behind us. In one other highway scenario while driving through a construction zone, the car neither tried to move left nor slow down as it approached and drove by construction and police vehicles, all of which had lights flashing. In many states, the law says move over or slow down, but FSD did neither.

On local suburban streets, FSD generally made terrible decisions in the following scenarios, all in full daylight, with excellent weather conditions.
  • Struggled a bit with our neighborhood’s only traffic circle. No cars in the circle. Stopped before entering, and then after entering failed to curve properly in a way that would keep it neatly in the traffic lane. On the way back, the loaner vehicle with a newer software version handled this empty traffic circle better, so that's nice.
  • Couldn’t pick one of two left turn lanes from one street to another when approaching a light. Both streets are divided, and as the car approaches the light, two left turn lanes are added on the left side. The car pulled toward the two left lanes and then randomly swerved back and forth between the lanes, never really committing to either one. So, I seized control and picked one. Fortunately, no cars behind me to get confused or report our car to the cops for such bad driving. White lines were fresh and clear - these are new streets, and for now the painted lines are quite visible. Weather conditions were perfect.
  • As we approached a 3-way intersection controlled by a light in the loaner car, it properly pulled into the left turn lane and prepared to make the left turn, but then right at the end, it suddenly did an early and dramatic left yank on the wheel for the turn, which could have put us into oncoming traffic or head-on into cars waiting patiently at the light so they could get their turn to enter the road we were on.
  • As we approached an unprotected right turn, the loaner car "California stopped" and tried to pull into oncoming traffic when it should have come to a full and complete stop, and waited for traffic to clear before turning right.
  • As we approached an unprotected left turn from an added left turn lane from a 2-lane road onto a divided street with four lanes and a median, the car suddenly pulled hard left and tried to drive right into the lane of cars closest to the median at the stop sign instead of staying to the right of the median. Very very not cool, and pretty darn scary.
So, yeah, technically, FSD tried to crash the car at least 2-3 times on this drive to and from the SC. And no, we'll not pay for FSD any time soon, as EAP mostly performs acceptably on roads where it's available for use and for now FSD seems to be quite a risk on local streets and roads.
I just do not see the value in EAP. Paying $6k for basically lane change capability on a 40k car seems unproportionate. Maybe if it at least stopped at red lights, you could argue it has more utility.

$12k for FSD can be justified IF you convince yourself that one day it will do what Tesla advertises. I know it's probably ways away and may not even be possible with HW3/4, but if you compare the features that do work today, FSD offers so much more utility bang for the buck.

As I said before, if you learn its limitations, treat it as L2 ADAS, and pay attention, FSD can be a great tool and very safe much of the time.
Even with all its flaws, I'm frequently able to complete a 20 mile commute to work without any disengagements.
 
  • Like
Reactions: pilotSteve
I have a Tesla 3 and the only reason I haven’t sold it is because I’ve been counting on a “final” release of FSD to Jack the selling price of the car. Am I delusional?
Ha. Talk to people around here: You've got answers of "when will FSD have a fork in it" of Never, Two Months, two quarters, and two years. Some of these people have Actual Expertise in NN's and such, but none of them are the people doing the down and dirty in California.

Musk, the main (but not only) source of Dates for the software has been endlessly enthusiastic; people joke a lot about Elon Time.

But FSD is, truly, a Research Project. Nobody has ever gotten close to what Tesla is attempting. Being a research project (research being defined as, "Running Up Alleys To See If They're Blind"), there's some chance that the answer will be No. My personal opinion is that it might take six months to a year; it all depends upon there being a breakthrough or two. And, mind you, there have been breakthroughs already. Compare the 9.x, 10.x, and 11.x versions of the software and there's a definite strong progression to more and more capable self-driving software.

If the 12.x branch, coming in anywhere from two to four months, is as capable as the X/Youtube videos show (and, believe you me, they were pretty blame impressive), I'd guess that there'd be at least two, maybe three major releases to go, cleaning up the various messy bits after the main functionality is up and running. So, maybe the end of next year?

But that's all guesswork. It'll be done when it's done.
 
I just do not see the value in EAP. Paying $6k for basically lane change capability on a 40k car seems unproportionate. Maybe if it at least stopped at red lights, you could argue it has more utility.

$12k for FSD can be justified IF you convince yourself that one day it will do what Tesla advertises. I know it's probably ways away and may not even be possible with HW3/4, but if you compare the features that do work today, FSD offers so much more utility bang for the buck.

As I said before, if you learn its limitations, treat it as L2 ADAS, and pay attention, FSD can be a great tool and very safe much of the time.
Even with all its flaws, I'm frequently able to complete a 20 mile commute to work without any disengagements.
I get your point about the unreasonable cost, but I will offer a minor correction: with EAP the car stops at traffic lights and stop signs, even when just using the most basic "cruise control" (speed control) feature. So, that is something. Of course, on those same local streets and roads EAP won't do FSD things, but based on my notes above regarding how FSD has behaved for me in the past 24 hours in that environment, I'm happy about that limitation. If and when FSD actually becomes useful in that environment, we can always just subscribe. OTOH, EAP is supposed to have auto-park, summon, and smart summon, and Tesla folks didn't tell me those features wouldn't be there or that the automatic wipers would be a hysterical mess until after I signed the deal and/or had the car.

I have a Tesla 3 and the only reason I haven’t sold it is because I’ve been counting on a “final” release of FSD to Jack the selling price of the car. Am I delusional?
Yep, pretty sure you're delusional on that count.
 
  • Funny
Reactions: pilotSteve
Ha. Talk to people around here: You've got answers of "when will FSD have a fork in it" of Never, Two Months, two quarters, and two years. Some of these people have Actual Expertise in NN's and such, but none of them are the people doing the down and dirty in California.

Musk, the main (but not only) source of Dates for the software has been endlessly enthusiastic; people joke a lot about Elon Time.

But FSD is, truly, a Research Project. Nobody has ever gotten close to what Tesla is attempting. Being a research project (research being defined as, "Running Up Alleys To See If They're Blind"), there's some chance that the answer will be No. My personal opinion is that it might take six months to a year; it all depends upon there being a breakthrough or two. And, mind you, there have been breakthroughs already. Compare the 9.x, 10.x, and 11.x versions of the software and there's a definite strong progression to more and more capable self-driving software.

If the 12.x branch, coming in anywhere from two to four months, is as capable as the X/Youtube videos show (and, believe you me, they were pretty blame impressive), I'd guess that there'd be at least two, maybe three major releases to go, cleaning up the various messy bits after the main functionality is up and running. So, maybe the end of next year?

But that's all guesswork. It'll be done when it's done.
I disagree on the status of FSD as a research project, because Tesla has been selling and marketing it as the real thing for quite a few years, and it's never been as good as the company said it was. As it stands today, FSD has been a gigantic scam for years, and I hope that someday the company and it's CEO are held accountable for that. Yeah, I know they use a ton of lawyers in shark-skined suits to tweak language in the sales contracts and behind the scenes, but what they keep marketing and selling is not what one actually gets, and that has been the case for far too long.
 
Last edited:
  • Like
Reactions: plumazul
I have a Tesla 3 and the only reason I haven’t sold it is because I’ve been counting on a “final” release of FSD to Jack the selling price of the car. Am I delusional?
Chances of that are low. I’m going to keep my ‘18 model 3 for my kids when they are ready to drive in 5 years. Not because Tesla will have L4 by then, but because FSD would be quite good by then.
 
  • Like
Reactions: Dracaris
All right. As far as the version numbers go: Near as anyone can tell, roughly around Xmas each year Tesla usually manages to get everybody on the same overall version number (like, 2022.12.x or something). Then, incidental fixes and improvements that don't involve FSD go out to everybody, …
Everybody???? I bought FSD and I’m locked in at version 2022.8.10.16
And the major releases like 2023.32.9 are going to everybody, and all of those have a software-unlockable version of FSD, currently 11.4.4.
 
Everybody???? I bought FSD and I’m locked in at version 2022.8.10.16
Hokay. Looked it up on Teslafi. Out of the 19,000 or so installs they monitor, there's, like 600 cars on this release.

I'm not even sure that it counts as FSD; they don't have the usual nn.mm.oo.pp numbering scheme (like, 11.4.7.2) for FSD labeled on it.

I think that the yyyy.ww.nn.mm on your release number implies that you have a 2022 release, dumped out the door initially on the 8th week in 2022. That's like February of 2022, although the "10.16" implies that they've been chugging along dumping out releases.

The current bunch of gonzos on the "active research" (my term, nobody else's) are on 2023.27.7 and have FSD-b 11.4.7.3. The people with relatively recent cars are on 2023.32.9 or a minor point release or two behind that. My SO's MY, of 2021 vintage, has EAP, no FSD, and is on 2023.32.9; my car, which is 9/2023 vintage M3, is on that same release but has a paid-for FSD of 11.4.4 installed. When I got it, it was on 2023.32.100.1 and had, like, 11.3.7.something (ugly), but got updated twice, first to 2023.32.7, then 2023.32.9, both of which had 11.4.4.

Let me look through your previous posts.. what version of FSD are you on? Hm. If you're sig is correct, you're talking about the 2017 Model S with AP3. Let me wander around TeslaFi a bit..

Hm. There are active installs to your version going on as we speak. And, interestingly, of the 600-odd cars (out of 19,000) that are on your release, four of them have gone on to install either 2023.32.9 or 2023.32.6 (both of which have 11.4.4) or 2023.26.12, which Teslafi doesn't indicate if it's an FSD load or not.

So, again, with feeling: What version of FSD are you running? Should be in the Release Notes or on the Software tab in the car.
 
So, again, with feeling: What version of FSD are you running? Should be in the Release Notes or on the Software tab in the car.
I paid for FSD on my S. I don’t currently have it. Never have. I’ve been told by a Tesla tech (ranger) that development on my car has stopped. I will NEVER have FSD. All I want from Tesla is a refund.
What’s happening to me will eventually happen to everyone. I’m just the canary.
 
I paid for FSD on my S. I don’t currently have it. Never have. I’ve been told by a Tesla tech (ranger) that development on my car has stopped. I will NEVER have FSD. All I want from Tesla is a refund.
What’s happening to me will eventually happen to everyone. I’m just the canary.
Like the thread over in the GB crowd: Sounds like Small Claims Court. They promised, they're not going to deliver. Do a bit of discovery. Maybe start with registered mail to the counsel for Tesla?