FSD Beta v11.x

S4WRXTTCS · Jan 29, 2022

Knightshade said:
I wonder if you can ignore stalk disengagements for this?

I'd think any "HOLY COW I NEED TO AVOID AN ACCIDENT" disengagements would be with brake or wheel right?

When I stalk disengage it's usually stuff like "There's people behind me, and I know FSD is gonna take WAY to long to turn right at this stop sign up ahead when it takes massively longer than needed to creep into an intersection with perfect visibility both ways and I don't want to piss off everyone behind me"

I'm almost 100% certain they don't count the stalk. I have no special knowledge, but I can't see it being a reflex movement.

So that leaves the brake, and the steering.

The steering is often what I use because its often a steering correction. In fact I wish steering didn't deactivate FSD Beta, and instead was using it as guidance. Things like I don't want to get this close to the edge or that I don't want to run over this pot hole. Or "the way you're turning is annoying me".

Sometimes a steering disengagement is purely accidental, but usually purposeful. Maybe 10% accidental disengagement.

I've been trying to use the brake to disengage as I hate how steering disengagement doesn't also turn off Adaptive Cruise Control.

There is some noise with the brake disengagement as well as sometimes I use the brake to turn off FSD Beta when someone pulls behind me at a stop light where I'm turning. I usually use the stalk as you described, but I hit the brakes sometimes.

I wish there was a giant red button with a USB port on it that could tell Tesla that I'm really annoyed, and you need to fix this problem. Where hitting it not only deactivated FSD Beta (if still activated), but also sent a report (like the tiny report button).

Basically the red button would serve as confirmation that it was a real valid disengagement over a critical issue that wasn't a map issue or nervous reflexes.

Included with the Red button would be 10 free map updates. Where all you had to do was email a special address your correction requests and they'd fix them.

sleepydoc · Jan 29, 2022

S4WRXTTCS said:
I'm almost 100% certain they don't count the stalk. I have no special knowledge, but I can't see it being a reflex movement.

So that leaves the brake, and the steering.

The steering is often what I use because its often a steering correction. In fact I wish steering didn't deactivate FSD Beta, and instead was using it as guidance. Things like I don't want to get this close to the edge or that I don't want to run over this pot hole. Or "the way you're turning is annoying me".

Sometimes a steering disengagement is purely accidental, but usually purposeful. Maybe 10% accidental disengagement.

I've been trying to use the brake to disengage as I hate how steering disengagement doesn't also turn off Adaptive Cruise Control.

There is some noise with the brake disengagement as well as sometimes I use the brake to turn off FSD Beta when someone pulls behind me at a stop light where I'm turning. I usually use the stalk as you described, but I hit the brakes sometimes.

I wish there was a giant red button with a USB port on it that could tell Tesla that I'm really annoyed, and you need to fix this problem. Where hitting it not only deactivated FSD Beta (if still activated), but also sent a report (like the tiny report button).

Basically the red button would serve as confirmation that it was a real valid disengagement over a critical issue that wasn't a map issue or nervous reflexes.

Included with the Red button would be 10 free map updates. Where all you had to do was email a special address your correction requests and they'd fix them.

I would agree - the biggest 'emergency' disengagements are when you're about to hit something and you don't take the time to hit the stalk, you just grab the wheel. Braking would fall under that category, too but is less common, IME.

The nice thing about using the stalk to disengage is it's smooth and doesn't jerk the car around like the steering wheel does.

momo3605 · Jan 29, 2022

sleepydoc said:
What's wrong with 10.9?

Jerkier turning/swerving and harder braking events for unexpected pedestrians and cars jumping in your lane.

jabloomf1230 · Jan 29, 2022

Knightshade said:
OTA updates?

Less than a year longer.

OTA was introduced in iOS 5, so the first update to that was November 10, 2011.

Tesla began doing OTA updates with the Model S in October 2012.

So both have been at this nearly as long, and for nearly a decade.

When Apple started doing OTA updates there were approximately 150 million iPhones that had been sold previously. When Tesla started doing OTA updates, it had sold less than a 1000 cars. The experiences are not comparable. Besides Apple's main concern was maintaining tight security. Apple tends to value security over profit margin. Maybe Tesla does updates to foil hackers, but we never see that mentioned in the release notes.

So you don't like my speculation. What's your evidence that supports an alternate hypothesis?

JulienW · Jan 30, 2022

jabloomf1230 said:
...So you don't like my speculation. What's your evidence that supports an alternate hypothesis?

Tesla has over 2,000,000 cars and all are getting their regular updates (sans a scant few Beta testers). So you are saying Tesla has enough server capacity to update the entire fleet unabated but can't spare any "extra" server capacity to even squeeze in ONE single update in the last week to an insignificant minority of Tesla's for Beta testing?

No way Tesla stoped the 10.9 update with just 20K or so more to complete because they ran out of server capacity for 10.9. If Tesla was having a server problem then all updates would be effected.

Also it is likely that just normal "under the hood" car to server to car data exchanges exceeds the data required for Updates making Updates even more insignificant to the servers.

EDIT: The same could be said for Apple too. Day to day iCloud, Apple Music and Apple TV and other services data will far exceed Updates also.

Knightshade · Jan 30, 2022

jabloomf1230 said:
When Apple started doing OTA updates there were approximately 150 million iPhones that had been sold previously. When Tesla started doing OTA updates, it had sold less than a 1000 cars. The experiences are not comparable.

That's true.

Tesla should have far more experience learning to scale and grow their OTA capacity since Apple already had to START big.

Not to mention as others have pointed out thanks to shadow mode and FSD data collection campaigns Tesla in general exchanges far more data with their cars than Apple does with their phones on a regular basis.

SW updates are only a tiny fraction compared to the amount of data FSD users are providing Tesla.

jabloomf1230 said:
Besides Apple's main concern was maintaining tight security. Apple tends to value security over profit margin

.....I guess you're unaware Apples profit margin is even larger than Teslas?

I'm not really seeing how that's relevant to server capacity though so this seems like a grasp at straws? Both have massively, massively, massively, more cash on hand than needed to add bandwidth if required for their servers.

jabloomf1230 said:
. Maybe Tesla does updates to foil hackers, but we never see that mentioned in the release notes.

Except yes, we have.

Most famously older S/X owners were forced to update OTA, or lose the ability to do so at all, specifically due to security updates.

All owners on older SW received this message specifically:

OTA note from Tesla said:
The Tesla network is undergoing enhancements for increased security. In order to maintain compatibility with and access to connected vehicle features, this vehicle requires a software update to at least version 2019.40.2.3. If not updated prior to 1-May-2020, this vehicle may no longer be able to receive over-the-air software updates, access the Tesla Mobile App & associated features, utilize voice commands, receive streaming media content, and other connectivity dependent features may be impacted. Please install the available software update by selecting the yellow clock icon and choosing a convenient time. If you are persistently experiencing software update installation failures, please schedule a service appointment via Tesla Mobile App.

That said- I still don't see the relevance to bandwidth for updates.

Not to mention since Teslas don't run 3rd party apps, and don't run any content from any source but Tesla outside of Tesla-provided apps (for music and video streaming and games, or in a sandboxed browser) remote security vulnerability vectors are far far lower.

jabloomf1230 said:
So you don't like my speculation. What's your evidence that supports an alternate hypothesis?

All of it?

There's been no noticeable change in update rate for anyone except the folks waiting on 10.9.

So the only evidence we have supports the idea THAT software, specifically, has an issue-- since it's the only one that appears to have basically stopped being pushed out among newer SW versions.

Teslafi currently shows 5.4% of fleet on 10.9, and 8.7% still on 10.8.1. Meaning only about 1/3rd (~38%) of Beta users got 10.9 at all before the rollout basically stopped.

Meanwhile rollouts of non-FSD versions of the same update as 10.9 continued in significant numbers all week. Making it clear it was the FSD part of that update that had a problem.

jabloomf1230 · Jan 30, 2022

I have a few questions that I haven't seen addressed. When Elon says V11 will be single stack, I've always assumed he meant FSD beta. But in the long run, what will happen to NOA in non-FSD cars? Are they stuck with the present approach? Or will some benefits of FSD single stack eventually bleed over to NOA? In my experience, FSD beta has reduced the number and severity of phantom braking events. It seems like NOA could also benefit from these improvements, considering phantom braking seems like a common complaint.

jabloomf1230 · Jan 30, 2022

Knightshade said:
So the only evidence we have supports the idea THAT software, specifically, has an issue-- since it's the only one that appears to have basically stopped being pushed out among newer SW versions.

Is it something that Tesla discovered, because none of the 10.9 videographers have complained about either any glaring bugs or consistent regressions? If Chuck Cook and guys like greentheonly can't uncover what's wrong with 10.9, then it probably was stopped for another reason.

This is going way off topic, so I'm not responding anymore to posts regarding why the release of 10.9 was stopped.

Twiglett · Jan 30, 2022

jabloomf1230 said:
I have a few questions that I haven't seen addressed. When Elon says V11 will be single stack, I've always assumed he meant FSD beta. But in the long run, what will happen to NOA in non-FSD cars? Are they stuck with the present approach? Or will some benefits of FSD single stack eventually bleed over to NOA? In my experience, FSD beta has reduced the number and severity of phantom braking events. It seems like NOA could also benefit from these improvements, considering phantom braking seems like a common complaint.

In the currently sold cars NOA is part of FSD, but a few years ago NOA was available as part of EAP.
So the only way to have NOA without FSD is to have an older car with EAP

That said, single stack implies that Tesla are planning on using the FSD stack of Autopilot for everything.
Remember the purpose of the single stack is to use it for all functions, so autopark, smart summon, dumb summon, AP and FSD rather than the current situation where they each have their own code.

On the 10.9 thing - I've seen several videos showing how 10.9 is now more prone to choosing the wrong lane leading to oncoming traffic problems. They reported it as happening in places that were previously fine.
It has gone oddly quiet with regard to new FSD versions, maybe they really are closer to giving up on 10.x and moving to 11.x

S4WRXTTCS · Jan 30, 2022

jabloomf1230 said:
I have a few questions that I haven't seen addressed. When Elon says V11 will be single stack, I've always assumed he meant FSD beta. But in the long run, what will happen to NOA in non-FSD cars? Are they stuck with the present approach? Or will some benefits of FSD single stack eventually bleed over to NOA? In my experience, FSD beta has reduced the number and severity of phantom braking events. It seems like NOA could also benefit from these improvements, considering phantom braking seems like a common complaint.

To my knowledge NoA doesn't exist to any major extent on cars that have HW3, but don't have FSD. The reason for this is Tesla removed the Enhanced Autopilot Option that has NoA, and moved it to FSD before they really started to ship vehicles with HW3.

To answer your question NoA in non-FSD cars are stuck with legacy code that predates pure vision since pure vision won't work on HW2.5.

The last version of firmware I used before pure vision seemed pretty good in terms of minimal phantom braking. So I don't think HW2/HW2.5 folks are necessarily left in a bad spot.

They do have a right to complain as their NoA, Autopark, and Smart Summon performance will forever kinda suck. When I say suck I don't mean everyone will have a bad experience, but that enough of them will be "this sucks" that it will never have a high level of satisfaction among owners.

Now this doesn't mean FSD owners will ever be okay with NoA, AutoPark, and Smart Summon. But, at least we have hope.

This kind of situation is exactly why I got FSD on top of EAP. It didn't make any sense to me to be left with older HW.

sleepydoc · Jan 30, 2022

Knightshade said:
That's true.

Tesla should have far more experience learning to scale and grow their OTA capacity since Apple already had to START big.

Not to mention as others have pointed out thanks to shadow mode and FSD data collection campaigns Tesla in general exchanges far more data with their cars than Apple does with their phones on a regular basis.

SW updates are only a tiny fraction compared to the amount of data FSD users are providing Tesla.

.....I guess you're unaware Apples profit margin is even larger than Teslas?

I'm not really seeing how that's relevant to server capacity though so this seems like a grasp at straws? Both have massively, massively, massively, more cash on hand than needed to add bandwidth if required for their servers.

Except yes, we have.

Most famously older S/X owners were forced to update OTA, or lose the ability to do so at all, specifically due to security updates.

All owners on older SW received this message specifically:

That said- I still don't see the relevance to bandwidth for updates.

Not to mention since Teslas don't run 3rd party apps, and don't run any content from any source but Tesla outside of Tesla-provided apps (for music and video streaming and games, or in a sandboxed browser) remote security vulnerability vectors are far far lower.

All of it?

There's been no noticeable change in update rate for anyone except the folks waiting on 10.9.

So the only evidence we have supports the idea THAT software, specifically, has an issue-- since it's the only one that appears to have basically stopped being pushed out among newer SW versions.

Teslafi currently shows 5.4% of fleet on 10.9, and 8.7% still on 10.8.1. Meaning only about 1/3rd (~38%) of Beta users got 10.9 at all before the rollout basically stopped.

Meanwhile rollouts of non-FSD versions of the same update as 10.9 continued in significant numbers all week. Making it clear it was the FSD part of that update that had a problem.

I’m general I agree, but not of what you said applies.

Apple actually exchanges quite a bit of data with their devices - they automatically create iCloud backups on a regular basis. They sync files, passwords, photos, mail and other information on a near continual basis. There have been times when Apple's servers are down and it causes significant issues. (Also, Apple does intermittently have issues with server capacity.)

I don’t know if Tesla runs their own servers or if they contract it out. Most companies use a service like AWS or MS Azure. If that’s the case then it’s just a matter of updating the contract. Actual capacity is not a concern. The other half is the LTE connection. I believe Tesla contracts with AT&T for this and there may be data or time limits in that contract. (This isn't an issue if you're connected to a wifi network, though.)

Comparing iPhone data usage to Tesla data usage is a Apples to oranges comparison (pun intended.) the majority of iPhone apps do not use Apple's servers; they simply use Wi-Fi or cellular data. Server capacity is irrelevant here.

Despite all of this, I agree; there's no real evidence that Tesla updates are limited by server capacity. I think it very plausible and likely that Tesla staggers and stages its upgrades. Doing so would have several advantages. Beyond evening out the server load, it also gives Tesla the chance to evaluate for bugs and other issues. In the event of a bug they can halt the rollout, patch the bug and continue with the next version.

EVNow · Jan 30, 2022

sleepydoc said:
I don’t know if Tesla runs their own servers or if they contract it out. Most companies use a service like AWS or MS Azure.

A lot of large tech companies first push out to network companies like Akamai and most users consume from their servers. Not even from AWS / Azure.

lakefiller · Jan 30, 2022

jabloomf1230 said:
Apple tends to value security over profit margin.

Apple still sends unencrypted vulnerable as hell SMS messages when iPhones need to message non iPhones. They could have switched to RCS by now which does have E2E for communication with non iPhones. They aren't doing it because it would hurt their profit margins and make iPhones less appealing because the iMessage lockin would lesson.

sleepydoc · Jan 31, 2022

lakefiller said:
Apple still sends unencrypted vulnerable as hell SMS messages when iPhones need to message non iPhones. They could have switched to RCS by now which does have E2E for communication with non iPhones. They aren't doing it because it would hurt their profit margins and make iPhones less appealing because the iMessage lockin would lesson.

RCS is not universal or standardized and if you’re worried about security (moderator edit). Secure messaging applications like Telegram have been available for years.

(moderator comment: evidence that it is becoming a universal standard here. A counter argument on RCS here.

N00bMaster69 · Jan 31, 2022

LastGas said:
Following too close only applies to speeds 50 mph and over. The score is based on a ratio of the time driving under 1 second to the time driving 1-3 seconds behind the other car. To get a good score you have to have a small number on the top and a big number on the bottom.

What screws you is Autopilot. Neither the bad driving nor the good driving accumulates when Autopilot is active. If you use Autopilot "all the time" like I would if I were driving the Interstate in Atlanta, the only time that counts is the scraps when Autopilot gets turned off, such as when you're making a difficult maneuver.

If you want a good following score, turn off Autopilot and let long stretches of normal driving add up in the divisor. As long as you can keep it over 1 second a fair amount of the time, you're cool.

I turned off Autopilot and my safety score jumped from all over the map to a consistent 100. When the average hit 97, I got the Beta.

This is both accurate and in accurate. I realized after literally driving hundreds of miles before I got FSD that this holds true for your daily score, it will only aggregate your actual manual miles together. But if say you drive 100 miles on AP and get a 100 daily score, that 100 is weighted based on the 100 miles driven, regardless of having AP on, against your last 30 days.

aronth5 · Jan 31, 2022

I take this as a hint from Whole Mars Catalog that we won't see a 10.9.x release which isn't a surprise. He is also being careful to say "might be" which may be to protect his source. His track record is very good

https://twitter.com/search?q=next big push might be&src=typeahead_click

NigelH · Jan 31, 2022

sleepydoc said:
How about all the times I disengage autopilot proactively because I know it can't handle a situation?

I had to take over twice in my 7 mile commute to work today. Both times it decided to enter a turn lane rather than stay in the main traffic lane. When it does this it will either keep driving on the shoulder or jerk back into the traffic lane, potentially causing a dangerous situation if there are other cars around.

If I pull the steering wheel it would register but likely not be viewed as a 'potential accident.' If I prophylactically disengage it would not be viewed as anything, but it's still a failure that likely won't be counted.

I'm guessing they are looking at the 3 forms of disengagement as @Mardak listed and doing a random sampling. Even with that it would be difficult to get an accurate measure.

I had the same thing happen to me twice on one drive after I loaded 10.9. Hasn't repeated it since. I put it down, that the car was having a bad day and or on drugs.

Knightshade · Feb 1, 2022

Sounds like Tesla has been discussing the rolling stop thing in aggressive with NHTSA since about Jan 20, so one theory is they stopped pushing new FSD versions around then because they knew they'd need to push one that removed that.

EVNow · Feb 1, 2022

aronth5 said:
I take this as a hint from Whole Mars Catalog that we won't see a 10.9.x release which isn't a surprise. He is also being careful to say "might be" which may be to protect his source. His track record is very good

View attachment 762732

https://twitter.com/search?q=next big push might be&src=typeahead_click

He definitely has some sources ... we just don't know whether those sources know these first hand or not. Also, Tesla can change plans quickly ...

I feel Elon would prefer 11 after 10.9. It just sounds better.

But the silence is likely because of the NHTSA forced recall.

sleepydoc · Feb 1, 2022

Knightshade said:
Sounds like Tesla has been discussing the rolling stop thing in aggressive with NHTSA since about Jan 20, so one theory is they stopped pushing new FSD versions around then because they knew they'd need to push one that removed that.

Possible - the ironic thing is coming to a full stop will probably cause more rear-endings because virtually no one stops completely unless there's a reason to.

FSD Beta v11.x

Well-Known Member

Well-Known Member

Active Member

Minister of Silly Walks

Well-Known Member

Well-Known Member

Minister of Silly Walks

Minister of Silly Walks

Single pedal driver

Well-Known Member

Well-Known Member

Well-Known Member

Member

Well-Known Member

Member

Long Time Follower

New Member

Well-Known Member

Well-Known Member

Well-Known Member

Similar threads