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.
Nice to see Elon's message that the separate highway stack will be going away fairly soon.
One has to assume they will focuse on fine tuning as they approach 8/8 so need to get rid of the big things.
There's already a FSD(Unsupervised) build in testing by Tesla - this is likely what will be officially announced on 8/8. My theory is that FSD(U)/FSDU will be the first L3 version. It won't be L4/L5. The manned robotaxi network will use the FSDU version - the driver is only there to take over when FSDU fails and needs human takeover in other words. Just my theory of course - only time will tell.
 
There's already a FSD(Unsupervised) build in testing by Tesla - this is likely what will be officially announced on 8/8. My theory is that FSD(U)/FSDU will be the first L3 version. It won't be L4/L5. The manned robotaxi network will use the FSDU version - the driver is only there to take over when FSDU fails and needs human takeover in other words. Just my theory of course - only time will tell.
I think you have a good chance of being right.
 
There's already a FSD(Unsupervised) build in testing by Tesla - this is likely what will be officially announced on 8/8. My theory is that FSD(U)/FSDU will be the first L3 version. It won't be L4/L5. The manned robotaxi network will use the FSDU version - the driver is only there to take over when FSDU fails and needs human takeover in other words. Just my theory of course - only time will tell.
Why not use remote assistance so you can have a single “driver” run many robotaxis?
 
Yep. They’ll likely be forced to do the NOA transition before then though.

Just seems unlikely that this will be ready enough for a wide FSD fleet release before v11 is ready for a fleet release.
I'm not sure that updating Autosteer/NOA is even in the plans. You want the better performance of FSD, subscribe to it...

The incentive will just be too strong. Of course concurrent limited release could still occur and still meet the mileage requirement.
What incentive? Of course, FSD will likely be safer, so maybe they will at some point, but I don't see any strong incentive at this point. (NHTSA seems more worried about keeping the driver engaged than on trying to demand that AP be perfect.)
 
What incentive?
Mileage as extensively discussed.

The graph is critical. They could stop publishing it of course - another possible outcome.

Slope will likely decrease otherwise.

FSD usage slope increases a lot with v11 highways. v11 resulted in big increase in slope recently, biggest contributor.

And as discussed, if safer eventually something will need to give there. Obviously a high bar there.
 
Last edited:
I think it's worth pointing out that while past experience matters, it doesn't matter as much within the context of an entirely new FSD v12 build that was a complete rewrite from the ground up that uses a completely different AI end-to-end approach (at least for city streets) when compared to all prior versions. So while experience with FSD use cases and other points certainly may help, some of the base assumptions have likely changed with FSD v12 and need to be challenged. In some cases, that past experience can actually serve as a roadblock to progress regarding how the new code base actually functions vs past FSD versions.

As someone who has managed personnel for over 20 years, whenever I bring in a new hire, I explicitly instruct them that I will meet with them at 30, 60, and 90 days, and my goal is to capture their fresh outside perspectives on what's working and what's not working. Too often, institutional knowledge from long time employees actually serves to prevent forward progress, so I encourage net new employees with a fresh outside perspective to be as open and honest as is possible - to challenge narratives and processes inbred into the company culture over time by those with the most institutional knowledge/past experience. It's a balancing act in other words - and I think this same concept applies at least to some extent when it comes to getting new and fresh perspectives on the new FSD v12 version.
Don't take most of what I post seriously, unless I am specifically speaking about a behavior relating to the car and the way the software is performing.
Those posts come infrequently, but are well grounded with miles of experience from V10.2 to current release.

While I agree with your overall thesis, at the same time people who were 100% safety score testers, such as myself and many others here, have had it soo long that we are able to detect the most trivial of changes in any performance of the software.

This is something, that someone starting on V12 does not have the expertise of knowing. This is due to the old software was always 2 steps forward, 1 step back, and those using the system at the time learned to be hyper aware of the progressions and regressions.

Since V12 came out, and all the subsequent releases, I have noticed no regressions from update to update. The only exception could be the ASSO addition, in which most users are pressing the go pedal when they need extra speed or disengaging when going to fast. Turning off the ASSO seems to have ZERO change as to how fast ego (the car) maintains speed.
Personally, for me, V12 is a quantum leap forward after suffering for years on all the other stacks.
 
Why not use remote assistance so you can have a single “driver” run many robotaxis?
I suppose that's feasible - but it would require a certain level of network connectivity to allow remote control and camera displays coming from the robotaxi - I also have no idea about the legalities/approvals that may be involved for this type of thing.
 
I suppose that's feasible - but it would require a certain level of network connectivity to allow remote control and camera displays coming from the robotaxi - I also have no idea about the legalities/approvals that may be involved for this type of thing.
Remote assistance can ONLY work with L4. L3 REQUIRES someone in the driver's seat ready to take over and physically and legally drive the the car.
 
Mileage as extensively discussed.

The graph is critical. They could stop publishing it of course - another possible outcome.

Slope will likely decrease otherwise.

FSD usage slope increases a lot with v11 highways. v11 resulted in big increase in slope recently, biggest contributor.
Why would it change? The people using FSD on the highway will continue to use FSD on the highway, even after it transitions to the FSD V12.5 stack.

You wouldn't put AP mileage into a FSD usage statistic, even if it used the same NN.
 
Don't take most of what I post seriously, unless I am specifically speaking about a behavior relating to the car and the way the software is performing.
Those posts come infrequently, but are well grounded with miles of experience from V10.2 to current release.

While I agree with your overall thesis, at the same time people who were 100% safety score testers, such as myself and many others here, have had it soo long that we are able to detect the most trivial of changes in any performance of the software.

This is something, that someone starting on V12 does not have the expertise of knowing. This is due to the old software was always 2 steps forward, 1 step back, and those using the system at the time learned to be hyper aware of the progressions and regressions.

Since V12 came out, and all the subsequent releases, I have noticed no regressions from update to update. The only exception could be the ASSO addition, in which most users are pressing the go pedal when they need extra speed or disengaging when going to fast. Turning off the ASSO seems to have ZERO change as to how fast ego (the car) maintains speed.
Personally, for me, V12 is a quantum leap forward after suffering for years on all the other stacks.
All good points. I've actually noticed an uptake in hitting the go pedal since I've always got ASSO enabled. I'm often hitting the go pedal - likely several times each trip - whether around town or a longer road trips - as FSD is still too timid/shy in many circumstances- especially when navigating through stop signs or a scenario where a rolling stop - such as a right turn on red where the signs indicate that stopping isn't required. I've not had an instance that I can think of where FSD is going too fast and I've actually disengaged - though I'll admit when I enable FSD exiting our driveway - and it tentatively exits the driveway then accelerates pretty swiftly - in a suburban development - it's really going too fast at that moment - but after a second or two it modulates the speed and corrects itself.
 
Remote assistance can ONLY work with L4. L3 REQUIRES someone in the driver's seat ready to take over and physically and legally drive the the car.
There are companies that claim it’s possible to safely remotely drive cars.
 
There are companies that claim it’s possible to safely remotely drive cars.
Not sure how that could be legal since that is the same as L0 driving. Seems if in an accident or moving violation the driver would need to be in the car since it is NOT autonomous and the same as any other car being driven by a person. How does the police see the driver's license for instance?