FSD now $8,000. Who's buying?

[aronth5]

Does anyone think even at this point that a tesla could drive itself from the west coast LA to new York without intervention. ?? wasn't that supposed to happen back in 2019/2020?

Without intervention not a chance.. Without disengagement unlikely but I bet you could do it with 5-10. Remember almost all driving would be on the highway and if you waited until V12 is available on highways you'd be in a much better position. I've driven for hours at a time on the highway many times without any interventions/disengagements.

Almost all of my disengagements are to provide Tesla with feedback. I rarely do safety critical disengagements now. Maybe one every couple of weeks. Then again I don't drive into Boston every week either.

 

[Joey D]

I hear this a lot as an excuse to try it out but I’ve gotta say, there’s just no world right now where FSD is a stress reducer on a long trip. It requires SO much more attention than basic autopilot.

Simple AP is a game changer for long drives - stay between the lanes, follow the car in front of me, no funny business. But as soon as the car starts trying to think for itself, it’s a whole different scenario.

Honest question: have you actually used the latest version of FSD on the highway? (Not EAP, but the latest FSD build which I believe uses something like the version 11.x FSD stack for highway driving.)

Because in my experience, there is no way that FSD on the highway is stress inducing. When I first used EAP (about a year ago...whatever code that was using....) I can see that argument, but not now.

 

[bdwalters]

I still think the subscription is the way to go versus $8000. I’m probably not going to keep the car long enough to make buying it worth it, and I would hate to total the car or trade it in earlier than expected and lose the full $8000. At $5000, I would start thinking about buying.

 

[Ben W]

My best guess is that Tesla is using the random "transferable FSD" option as a way to control not sabotaging profits if they were to apply FSD to the Account instead of the VIN.

Tesla will always figure out a way to get what they want at the expense of anything. There are no emotions attached to Tesla in what they do, with the exception of maybe keeping the S and X models alive due to sentimental reasons, but for everything else it's just business. Like trading, we just need to study and make the right entry point. It's all a game.

I'm also in the market for a S and have dabbled into looking at Lucid. The biggest culprit for me with Lucid is the horrendous CCS port (soon to change) but also to some degree the Software UI. Tesla has by far the best software and it's a big determining factor for a lot of people, despite the other flaws and inferiority that Tesla posses over actual luxury cars like Lucid.

Fun fact: Lucid has discounted some of their in stock inventory up to ~$30K or something insane like that. A lot have been sold, but there may be some outliers left.

I think in a few years, if not a lot sooner than we think, we'll start to see some surprises from other OEMs with their advancements in Autonomy. The EAP like systems in other makers right now are really good and in some ways better than ours. When it comes to FSD, this is where Tesla is rolling in the bank of data, but that doesn't mean that other OEMs aren't working on their own tech as we speak.

Although they all, including Telsa, have flaws to some degree. watch out for Mercedes and HMG products. They have excellent driving assistance in their own way that top the rest of the legacy makers. Ford Blue drive isn't bad either.

Exciting times...

Agreed that it's "all a game", and just a question of finding the right time to strike. I'm not in a huge hurry; eventually there will be HW5, bidirectional charging, improved battery chemistry, an exterior refresh, and so forth. I'd be very surprised if they do away with the S and X altogether (and lose "S3XY"? Never!); more likely they will eventually get a complete refresh onto a state-of-the-art platform (making them much easier to manufacture), though Tesla has to dig its way out of its current situation first.

As far as Lucid goes, I wouldn't be surprised if they find a way to make NACS retrofittable to replace the CCS port on older cars. (Gruber Motors is currently working on a NACS retrofit for the original Tesla Roadster, even.) Software can be updated, so the Lucid's mediocre UI is not a showstopper; no doubt it will continue to improve. (I haven't taken it for a test drive yet, but plan to.)

Tesla may be rolling in training data, but it's only pure-vision data, so it may not be adaptable to train other systems that employ radar and/or lidar. This may pose a serious impediment, if and when other manufacturers eventually decide to "license Tesla FSD software" while keeping their own sensor hardware. I suspect that Tesla's somewhat brute-force training approach may eventually be leapfrogged by other companies who learn how to do more with less; it really shouldn't take 10,000 years of training clips to properly train a driving system! Humans are capable of learning quite a lot from a single example; neural nets should eventually be, too.

Looking forward to see what the future brings, exciting times indeed!

 

[Ben W]

For me 12.3.4 can do it [coast to coast drive with no interventions] IMO. Maybe my cameras are better calculated than the bears on this forum

"Can", if it's extremely lucky. It's technically accurate to say that properly calibrated 12.3.4 has a nonzero probability of completing a coast-to-coast drive successfully and safely, with zero interventions. (I'd guess the probability of success on a given attempt would be less than 1%.) To be L4-ready, or even L3-ready, this 1% needs to improve to 99.99%. (~30 million miles between unforced accident-causing errors or failures.) There is a vast, vast gulf between L2 and L4, and Tesla is just barely beginning to tiptoe across it.

 

[ucmndd]

Honest question: have you actually used the latest version of FSD on the highway?

My last FSD experience was late 2023 when I got a 3 month trial with my Model 3.

Maybe the new versions are better. Probably even. But they’d have to be WAY better for me to want to waste my time. I’ve been using autopilot-enabled Tesla’s since 2016 and fell for the Elon snake oil all the way back then.

I dont trust it. It’s given me way too many reasons not to. My version of driver assistance in 2024 is still “stay where I put you until I tell you to do something else.” Basic autopilot is a wonderful tool for commuting and I drive 35k+ miles a year.

I can’t even get Full Self Wiping to work and not constantly smear bug guts all over my perfectly dry windshield. Driving? lol, let’s solve the windshield wipers first.

 

[Tam]

Honest question: have you actually used the latest version of FSD on the highway? (Not EAP, but the latest FSD build which I believe uses something like the version 11.x FSD stack for highway driving.)

Because in my experience, there is no way that FSD on the highway is stress inducing. When I first used EAP (about a year ago...whatever code that was using....) I can see that argument, but not now.

EAP/Autopilot are fine but the current FSD on highway is still stressful for me: It keeps doing unintentional Autolane changes even with "Minimal Lane Changes" setting.

Thus, I shut FSD off on highways and use "Autosteer" mode instead.

 

[Rocket_man]

Maybe there should be three FSD products...

FSD Lifetime. For an individual owner for one car, fully transferable for the lifetime of the owner to future Tesla cars. For personal use only, no ridesharing functions. This might be worth $5K, maybe more, and make buyers really brand loyal. I think I owned Honda products for 20 years of my car life. So $5K or even $8K would be worth it over that kind of lifespan.

FSD Monthly. For one-time Tesla owners, or those that do mostly short range driving but occasional long distance driving. $99/mo or less is probably the price range for this.

FSD Robotaxi. FSD plus ride share summons, billing, planning, rider monitoring, maybe includes Tesla Insurance. For operators of Robotaxi fleets or individuals that want to ride share a personal car. This could be priced higher if it makes your car a revenue generator but not sure what that price is. Not $100K, it would have to be much less. There are many liability issues to work out here and government regulatory issues that could change the business model overnight. This Robotaxi thing is one 'Robotaxi runs over Granny pushing stroller' accident away from being regulated out of existence.

 

[Joey D]

My last FSD experience was late 2023 when I got a 3 month trial with my Model 3....

Like I said, I can understand your view prior to the current stack. My experience in 2023 was similar. It's a different experience now.
I will have to see what EAP is like now (without the FSD stack) to see what differences may be

 

[JulienW]

......FSD Robotaxi. FSD plus ride share summons, billing, planning, rider monitoring, maybe includes Tesla Insurance. For operators of Robotaxi fleets or individuals that want to ride share a personal car. This could be priced higher if it makes your car a revenue generator but not sure what that price is. Not $100K, it would have to be much less. There are many liability issues to work out here and government regulatory issues that could change the business model overnight. This Robotaxi thing is one 'Robotaxi runs over Granny pushing stroller' accident away from being regulated out of existence.

Not happening. Tesla is building a detected robotaxie vehicle with a segregated data center (not using current cars the way they said in 20). Our cars will likely stay L2 with maybe some L3 highway thrown in.

 

[Joey D]

EAP/Autopilot are fine but the current FSD on highway is still stressful for me: It keeps doing unintentional Autolane changes even with "Minimal Lane Changes" setting.

It's not clear if you are talking about operating on a "limited access highway", or "Multi-lane highway like road." Those are different situations, and running on different stacks. Running FSD 12 on a "highway like road" may do more lane changes than you normally would.

I'm talking specifically about the following:
1) Latest FSD Beta Installed
2) Operating on a limited access highway (think "Interstate")...which to my knowledge does not yet use FSD 12....it uses some other FSD version curated for limited access highways (it may be some version of FSD 11).

My lane changes in these situations are of course not always exactly how I would drive myself...but it's actually pretty close...and more importantly always safe. I have never had to intervene a lane change because it wasn't being done safely.

Further, if I want to do a lane change that the car is not doing on its own, it's great to be able to tap the blinker, and know that the car will then change lanes itself...but only when safe (in case I missed someone in my blind spot).

 

[Joey D]

Not happening. Tesla is building a detected robotaxie vehicle with a segregated data center (not using current cars the way they said in 20). Our cars will likely stay L2 with maybe some L3 highway thrown in.

While this is possible, no way to know this yet. Hopefully, 8/8 will shed light on the broad scope / picture that Tesla has in mind for dedicated Robotaxis vs. autonomy for personal automobiles.

 

[Rottyman]

I bought my MP3 June 2023 with EAP. $6000 at the time. Now, I can upgrade to FSD for $2000. I'm not sure if I'm distracted by the discounted price and the real need to have it. I've been using the FSD for since Tesla gave us the 'free' month and I like it. It's a little less aggressive than the way I drive, but for $2000, it's nice feature. Debating the pros/cons. I plan on keeping the car for at least 3-4 years.

 

[kjlee80]

So for tesla with HW2.5, FSD purchase(not subscription) would only upgrade to HW3.0 for free, but no side camera upgrades?

 

[beatle]

Honest question: have you actually used the latest version of FSD on the highway? (Not EAP, but the latest FSD build which I believe uses something like the version 11.x FSD stack for highway driving.)

Because in my experience, there is no way that FSD on the highway is stress inducing. When I first used EAP (about a year ago...whatever code that was using....) I can see that argument, but not now.

I tried it out during the trial and disabled it. The following distance is too far and people think you're dawdling, then they cut you off or attempt to pass you on the right. This is with the distance set to minimum and most aggressive. EAP is better at following, but AP1 was better still. Not sure if FSD fixes the occasional swerve back into the original lane when changing lanes. That was a pretty rare event though, only 2-3 times a year, but AP1 never did that.

 

[mborkow]

I tried it out during the trial and disabled it. The following distance is too far and people think you're dawdling, then they cut you off or attempt to pass you on the right. This is with the distance set to minimum and most aggressive. EAP is better at following, but AP1 was better still. Not sure if FSD fixes the occasional swerve back into the original lane when changing lanes. That was a pretty rare event though, only 2-3 times a year, but AP1 never did that.

This weekend, with the latest stack enabled, my car made what I considered to be an unnecessary lane change to the left (into the fast lane) and then proceed to go slower than the pace of traffic in that lane (effectively cutting off the car that was behind me). That driver then lane changed right, sped up, and lane changed left to get back in front of me. I’m sure that driver was annoyed at me…I was annoyed at FSD for what I considered very poor etiquette all for a lane change that I wouldn’t have made (or if I had made would have committed to match the pace of traffic in that lane).

 

[Ben W]

Maybe there should be three FSD products...

FSD Lifetime. For an individual owner for one car, fully transferable for the lifetime of the owner to future Tesla cars. For personal use only, no ridesharing functions. This might be worth $5K, maybe more, and make buyers really brand loyal. I think I owned Honda products for 20 years of my car life. So $5K or even $8K would be worth it over that kind of lifespan.

FSD Monthly. For one-time Tesla owners, or those that do mostly short range driving but occasional long distance driving. $99/mo or less is probably the price range for this.

FSD Robotaxi. FSD plus ride share summons, billing, planning, rider monitoring, maybe includes Tesla Insurance. For operators of Robotaxi fleets or individuals that want to ride share a personal car. This could be priced higher if it makes your car a revenue generator but not sure what that price is. Not $100K, it would have to be much less. There are many liability issues to work out here and government regulatory issues that could change the business model overnight. This Robotaxi thing is one 'Robotaxi runs over Granny pushing stroller' accident away from being regulated out of existence.

I fully expect that "Model 2" and "Robotaxi" will be exactly the same hardware product, introduced on 8/8. The $25k price will be subsidized by the purchaser pledging a substantial cut of future Robotaxi revenue to Tesla. (Maybe there will be an option to pay a higher price upfront and keep more Robotaxi revenue.) The autonomous sensor suite will be further enhanced over HW4 (maybe they'll call it HW4+ or HW4.5), and the car will initially ship with the HW4 computer but with planned upgradeability to the HW5 computer when it comes. [As my 2017 Model 3 shipped with the HW2.5 computer but was later upgraded to HW3.]

I still don't expect full L4 Robotaxi autonomy until 2030 at the earliest, and it's hard to see how they can make the Robotaxi model work at all as L3, since the car will still need to reliably drive itself to go pick up the passenger, even if the passenger can then function as an L3 driver. But maybe they can contrive a way, starting geofenced or some such.

In general though, agreed on these three FSD flavors. Transferability of fully-owned FSD will be really important; it very likely will make the difference whether my next car is a Tesla or not.

 

[JB47394]

Easy money he's just leaving on the table.

Reducing the price by 10x and increasing the take rate by 10x produces no net monetary change, while leaving Tesla with a greater potential for people to use the system irresponsibly. Keeping the price moderately high encourages use by those who have a real interest in the system and are more likely to use it properly.

I was annoyed at FSD for what I considered very poor etiquette.

Fitting in with the flow of traffic is the next major hurdle for FSD. Today, it can manage somewhat, but there are plenty of causes for complaint.

If you have it on, I'd suggest turning off Auto Set Speed Offset. In my opinion, the old way works better (going by the posted speed limit plus your selected offset). Another setting to try is Minimal Lane Changes, which I turn on each time I enable or reenable the system. That prevents it from making "intelligent" lane change selections such as passing slower traffic and such. That setting only applies to highway driving.

 

[JB47394]

I fully expect that "Model 2" and "Robotaxi" will be exactly the same hardware product, introduced on 8/8.

I expect them to be built from the same platform, but I think the Robotaxi will be amped-up on sensors and compute power. It obviously would also lack driving controls and should have auto-closing doors. The Model 2 should be that same platform set up like today's Model 3. Because of the need to operate with and without driving controls, I expect steer-by-wire.

 

[Tam]

So for tesla with HW2.5, FSD purchase(not subscription) would only upgrade to HW3.0 for free, but no side camera upgrades?

Correct. So far, that's how it works unless there's a new policy.