Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

FSD Beta 10.69

This site may earn commission on affiliate links.
Just watch the regen bar. It tells you everything. When it is long, you do not want to see it jump to short, then to long again. When it is very short, fluctuations matter less.

In this video, you’d see the regen bar get long three times:
1) Response to lead vehicle slowing, and maybe the light (probably not).
2) Initial slowdown transient. But then you see it jump to get short again (this is bad and is lost opportunity to use regen!)
3) Final slowdown bar gets long again. Then gradually shortens to nothing.

What should have happened is one elongation of the bar to a fairly long but perhaps not max length, and then stay that way for a bit, then gradually shorten to nothing.

The rapid changes in the length will directly tell you how lacking in smoothness the stop is. You can do this in any FSD Beta video and “feel” the drive. An initial large jump is not necessarily bad, but the more gradually it goes to max length, the better. You just don’t want to see the long-short-long pattern. We’re not trying to signal SOS here.

Deceleration is mandatory, but high jerk is not.

Those HDMI hacks are definitely helpful for this!
I don't think the regen bar tells you everything about the whole process of braking!

I have no idea how accurate this article explained the process, but it sounds like a complicated process beyond my knowledge. What is Regenerative Braking?
 
  • Like
Reactions: sleepydoc
Aren’t user counts required to determine average usage per user?
Yup. What does that have to do with my point?

You were the one who brought up the idea that the current batch of testers people would be using 10.2 just as much as the current Beta. That was not related to the point Daniel And I were discussing. (I thought the thing you brought up was silly and offered a thought experiment as to why Tesla has rolled out so slowly idle your ideal the first beta we’re as capable and safe as the current.)

Daniel’s plots looked pretty flat when user counts were stable. But it’s a little hard to be definitive without dividing the plot by the number of users at each point.
There was a relatively flat point during the period that no beta user were added.

I cannot fathom how you think that’s related to my point. 🤷‍♂️

Don’t you get that there’s a correlation between capability and stability of the releases and the addition of new batches of drivers (with significantly lower safety scores)?

A flat point in users correlating to a flat point in miles driven would be entirely expected whether or not the original graph I shared illustrates that FSD has improved in capability while only showing miles driven.
I think they probably wanted to improve the driver monitoring. That’s clearly been the biggest shift since 10.2 - huge changes in how monitoring is done (accompanied by strike resets as they iterate).
Explain how this theory holds any water with the careful rollout of 10.69.x (and its significant new core component in the Occupancy Network). No changes to driver monitoring and even a strike reset.

Doesn’t hold a drop of water to me. 🤷‍♂️

Obviously there have been other improvements too, but I’m not sure they would greatly increase the usage, at least in the current beta group.

I am utterly, utterly bewildered how you think that is remotely related to my point.

They also probably wanted to get a sense for how much people would use it and in what scenarios, and their degree of attention, before taking a risk on a larger group.
This makes no sense.

So they gave it to about 1000 people for months. And during all that time just checking to see where they might use it and if they would abuse it?

They kicked people out right away for abuse. They got lots of novel traffic situations driven in. They had a process to react to these issues in place from the start. What were they waiting for to widen?

What makes no sense with your theory is if the version that went to 50k was not more capable and safe, why did they wait so long until that version to widen to 99 Safety? And again to 100k at 93 safety. And months again to 80 safety?

You’re not explaining that bit.

It is very hard to explain if your dogma is based on the fact that FSD hasn’t gotten more capable or safe since 10.2.

But very easy to explain if you believe what the vast majority of testers are reporting. That it had been getting better.
 
Last edited:
That's quite literally what people do when waiting to make a left turn from the main road.

It’s different.

1) Obtaining pose is much harder. To be equivalent, you must be parallel. This is much harder to do if you’re coming from a position perpendicular to that, and you have to make the shift at 20mph with limited space.

2) The maneuver you need make is different. You have to merge with fast-moving traffic that you have to use your mirrors or crane your neck to see (rather than just drive across three lanes of traffic). The car has eyes in the exact right direction, but then you have to trust it.

As I said, when the median is wide enough, or long enough to allow straightening with a dedicated area just for you, this strategy seems fine.

But at the moment, given the issues with localization consistency, I think it’s a bit risky in tighter situations. And I think we may see that strategy cause an accident at some point, because it is a departure from the way most people drive.
 
You were the one who brought up the idea that the current batch of testers people would be using 10.2 just as much as the current Beta. That was not related to the point Daniel And I were discussing

.I cannot fathom how you think that’s related to my point.
🤷‍♂️


Perhaps you could summarize what was your point. This started with a discussion about what the graph showed and what that meant.
Don’t you get that there’s a correlation between capability and stability of the releases and the addition of new batches of drivers (with significantly lower safety scores)?


Certainly that is what Tesla has been suggesting.
No changes to driver monitoring
How do you know?
even a strike reset.
There were a lot of people whining at Elon. I think that is why we got that (very bizarre!).
I am utterly, utterly bewildered how you think that is remotely related to my point
Again, hard to know exactly what point you were making with the graph.

I went back and reviewed your posts, and they plainly state that you think individual users are driving substantially more per month with FSDb than they used to. (Miles/user/month) Seems like we should figure out if that is true. I would guess it is about the same (especially after correcting for overall miles driven, due to pandemic easing, etc.).

I don't think the regen bar tells you everything about the whole process of braking!
It really does! The bar now shows both regen and friction braking, scaled properly by power (Tesla can calculate how much stopping power the friction brakes are applying). Even in most of the bad stops we see, the friction component is small though (not nonexistent, which is annoying!). I don’t think the scale is linear, particularly for very small power displays, but that doesn’t matter when using it for this assessment. It’ll even show higher values for a given deceleration on downhill, vs. an uphill.
 
Last edited:
Perhaps you could summarize what was your point. This started with a discussion about what the graph showed and what that meant.

That is the point I’m referring to. At first I thought you didn’t understand the point I was trying to make (without necessarily agreeing with it).

But I’ve been increasingly getting the impression that you wouldn’t even be able to quote the point of mine I’m making.

The former would be on me for not being able to explain. The latter… that’s where I think your contrariness is what I’m talking to, not a good-faith discussion of any point I’m making.

Certainly that is what Tesla has been suggesting.

Ah so Tesla is saying they are being cautious and waiting for the capability and safety to suit a wider and lower safety score tester pool, but your theory is that their true motives have actually been your theories?

This sounds like conspiracy talk.

How do you know?

So your theory is now that Tesla was improving driver monitoring when they said they were, but during the months between when they stated they were improving monitoring (months when there were no major outbreaks of attention lapse issues), during those silent months they were STILL working on driver monitoring, and THAT was the reason for the delayed expansion?

And all of this slow rollout played out again in a matter of a few weeks with 10.69, NOT because the first release of the new core tech (Occupancy Network) was rolled out slowly for safety (like Tesla said), but because they had somehow had a major regression in… wait for it… driver monitoring?

And you’re saying “how do I know” this slow 10.69.x rollout wasn’t because they were secretly working on monitoring? (To what end I wonder as there have been no major issues come out about monitoring, and no accidents).

Man this is straight up conspiracy talk.

There were a lot of people whining at Elon. I think that is why we got that (very bizarre!).

So again your theory is that the slow rollout is because Tesla is SUPER concerned about driver monitoring and attentiveness.

AND you think Tesla just allowed people back in to this version who had been kicked out by this often secret monitoring system advances? Simply because they complained?

That makes literally zero sense.

Again, hard to know exactly what point you were making with the graph.

Again. I’m not sure you can even QUOTE the point you’re arguing. It seems you just like to take a counter position (and your little passive-aggressive laugh reactions) to any post that suggests FSD might be improving.

I went back and reviewed your posts, and they plainly state that you think individual users are driving substantially more with FSDb than they used to. Seems like we should figure out if that is true.

I believe this true. Because I’ve seen dozens and dozens of people reporting it.

Has nothing to do with the point in question. The miles per person per release has nothing to do with the point I made about that chart.

If I recall, it was brought up as an obvious counter to your point where you suggested (and continue to go to increasingly incredulous lengths, see above) that FSD Beta is no better today than 10.2.



I’m just going to leave this here.

I like to believe that Daniel was discussing the issue in good faith. I have trouble believing the same of you any longer.
 
Last edited:
  • Like
Reactions: FSDtester#1
They are lying and you know the real score that they would be using a way less capable version just as much? 🤔
You might be overestimating how many more miles people are using FSD Beta with more recent versions than before. At least for my own behavior, even before FSD Beta, I would use "highway" Autosteer on city streets for probably 95%+ of my miles as the only places I couldn't use it is on was unmarked residential streets and maybe a hundred feet for each turn or occasional "difficult" intersection/situation. So if we say FSD Beta is basically doing 100% of my miles now, that's around a 5% increase in mileage rate compared to pre-beta. But most of that gain happened when I got 10.2, so 10.69 barely moved the needle.

But I'm probably not the average FSD Beta user where yes indeed people choose not to turn it on for city streets. Do these people try out 10.69 and see it's good enough to keep active for longer or do they encounter similar problems they've seen in past versions and wait for the next version again?

Then again, the average FSD Beta miles per tester is probably not even 3 miles a day, so there is plenty of growth opportunity to raise that closer to average miles traveled each day.
 
  • Like
Reactions: AlanSubie4Life
You might be overestimating how many more miles people are using FSD Beta with more recent versions than before.

The point in question is unrelated to this. Whether it is exactly the same miles per person per release, or whether it has been increasing (as I believe it has and many dozens have reported it has, especially with 10.69.x).

But I'm probably not the average FSD Beta user where yes indeed people choose not to turn it on for city streets. Do these people try out 10.69 and see it's good enough to keep active for longer or do they encounter similar problems they've seen in past versions and wait for the next version again?

Okay here just basic logic:

If you believe these average beta users exist (you seem to, you even call them average users)…

And if you believe they would be turned off by a version with problems exceeding their own personal tolerance threshold (again you seem to, and I tend to agree this would be the reaction of a decent number)…

And if you believe that FSD Beta has been improving…

Then you should believe that the average miles per person is increasing.

(If you don’t believe FSD is improving then you believe the average would be increasing as they do improve it.)

Now none of this really relates to that point I made about the context of that graph SHOWING cumulative miles driven illustrating that FSD Beta has indeed been improving in capability.

(Just like a graph showing only male/female birth ratios, if you understand the right context, can illustrate something about a culture’s changing approach to children and gender.)
 
But I’ve been increasingly getting the impression that you wouldn’t even be able to quote the point of mine I’m making.

I think your point is that since total miles driven are increasing at a much more rapid pace now, without accidents, with an overall less safe group of users, that means that FSDb must be much more capable and safer than it was initially. But you’ll probably say that was not your point and I should guess again.

I actually do want to understand the argument you’re making.

Ah so Tesla is saying they are being cautious and waiting for the capability and safety to suit a wider and lower safety score tester pool, but your theory is that their true motives have actually been your theories?
No. I have said time and time again that Tesla is interested in safety - and I even suggested they likely have improved safety of FSD, in the post above.
I’m not sure you can even QUOTE the point you’re arguing. It seems you just like to take a counter position (and your little passive-aggressive laugh reactions) to any post that suggests FSD might be improving.
I can’t count the number of times I have said that FSD Beta has improved. It’s certainly improved from the original release. I don’t know whether I would call it “much” better, but that is a subjective assessment. In very specific ways, I would say it is definitely much better (for example, at turning). Overall, however, as far as utility goes, I would say it is only slightly more useful, but still not very useful (for the actual urban surface streets driving use case). The utility was always decent in other contexts, but there it becomes less and less differentiated from AP.

I believe this true. Because I’ve seen dozens and dozens of people reporting it.
It would be good to see whether it is though.
 
Last edited:
You might be overestimating how many more miles people are using FSD Beta with more recent versions than before. At least for my own behavior, even before FSD Beta, I would use "highway" Autosteer on city streets for probably 95%+ of my miles as the only places I couldn't use it is on was unmarked residential streets and maybe a hundred feet for each turn or occasional "difficult" intersection/situation. So if we say FSD Beta is basically doing 100% of my miles now, that's around a 5% increase in mileage rate compared to pre-beta. But most of that gain happened when I got 10.2, so 10.69 barely moved the needle.

But I'm probably not the average FSD Beta user where yes indeed people choose not to turn it on for city streets. Do these people try out 10.69 and see it's good enough to keep active for longer or do they encounter similar problems they've seen in past versions and wait for the next version again?

Then again, the average FSD Beta miles per tester is probably not even 3 miles a day, so there is plenty of growth opportunity to raise that closer to average miles traveled each day.
Anecdote, not data. I'm definitely using FSDbeta more with the most recent release. I was already using Navigate on Autopilot pretty consistently on highways and freeways but not on surface streets. When I got FSDbeta I was very skeptical that there would any utility to it, it was just going to be fun experiment. I was surprised that I was finding utility in it (even with first releases). Usage was still quite limited because the early releases had big gaps in capacilities.

The current release "feels" like it has moved the needle. There are have several trips where it basically did all the driving. But, I also have encountered failure cases that I wasn't seeing before but in many of those cases it was simply because the car never got the chance before. One issue that I haven't noticed discussed yet (perhaps I missed it in the thread drift) is incorrect speed limits as the car turns onto a new road. It almost seems like it sometimes "forgets" to consult "the database" and it just uses the speed limit it last had until it encounters a new speed limit sign. This gets mildly exciting when the car turns from a 50 mph express to 35 mph street and it thinks it should be going 50. Also annoying when the speed limit increases and the car doesn't accelerate until it gets to the new speed limit sign when clearly the street has a higher limit than where it was before.

Other issues that others have mentioned that I've encountered; completely wrong lane selection (for turns), late/jerky turn lane selection, ignoring speed bumps.
 
But I'm probably not the average FSD Beta user where yes indeed people choose not to turn it on for city streets. Do these people try out 10.69 and see it's good enough to keep active for longer or do they encounter similar problems they've seen in past versions and wait for the next version again?
Yes, I am very curious what the plot would show about prior versions and whether it would show any evidence of decay of miles (per month per user) driven between versions. There may not be enough time between versions but I wonder about 10.12 since it was kind of a long period. But we’ll probably never know.
 
Then you should believe that the average miles per person is increasing
Yes, I do. However, I would guess very little of the growth of cumulative miles is attributable to the increase of capability of FSD Beta versions. How many more miles on average do you think is increased compared to say 10.2? 10% 50% 100% 500%?

The monthly rate of cumulative FSD Beta miles in March through May was around 4 million miles, and that monthly rate increased to roughly 7 million miles in June. FSD Beta population was roughly 60k before the most recent enrollment batch in June increasing to about 100k testers. Suspiciously 60k -> 100k is roughly a 70% increase in population while 4M -> 7M miles/month is around 75% increase. Both 10.12 increasing capability over 10.11 as well as Tesla adding more testers contributed towards the increased monthly mileage, but I would guess most of that gain came from adding new testers.

We'll probably see a similar pattern with 10.69.2.1 getting new testers (probably?) in the same month that 10.69 was released to the existing 100k testers, so there will be a mixing of signals that won't clearly show the capability vs test population effects for September's data point.
 
I think your point is that since total miles driven are increasing at a much more rapid pace now, without accidents, with an overall less safe group of users, that means that FSDb must be much more capable and safer than it was initially.

Bingo. This is exactly my point. Apologies that I cast aspersions about your motives.

(Well, I didn’t actually say that the pace is increasing much more rapidly (Only up to June, mind) only that the miles have increased very rapidly and at an increasing pace. But ultimately that’s semantics.)

What I fail to see is why you and others think miles driven per user per release matters to that. I believe it is increasing, but the point above stands if it is exactly flat.

You could not have gotten that number of miles with 10.2 with those people, in those places, in those timeframes with only one bollard boop.

You maybe could have if every driver were a highly dedicated test pilot such as in this group. But a LOT of the 93+ Safety folks got in with little or no knowledge of the program, the system limits, etc. 99 takes dedication, understanding of the criteria, and ability to control the vehicle to achieve for most home driving areas. 93 takes very little of that. 80 takes practically none at all, filtering out only those who have real trouble demonstrating care and control in achieving the arbitrary Safety Score.

Hell the original graph was posted not just to illustrate the improvement since 10.2, but to illustrate the rapid pace of capability of FSD since the architectural rewrite using literal new-frontier class of software development.

Nobody outside Tesla’s testers (who had maximum level of system knowledge) had driven a mile on FSD until relatively very recently. And now (June actually) we’re here at a gobsmacking number of miles.

With a crazily WIDE 80+ safety release just around the corner (because it is so much more capable and safe, while still being very much a WIP.)

Understanding who is now driving those miles (minimal to sometimes zero system knowledge behind the release notes, many not even that much), and how well it’s been going (where it counts most of all, crashes) tells you all you need to know about how this ain’t your uncle’s “FSD has been a year away for 10 years.”

It is lightyears beyond where it was a few years ago. With a few AU to go. 😉
 
Yes, I do. However, I would guess very little of the growth of cumulative miles is attributable to the increase of capability of FSD Beta versions. How many more miles on average do you think is increased compared to say 10.2? 10% 50% 100% 500%?
The error you’re making is conflating miles driven per person with miles driven. They are VERY closely linked. But miles driven per person is not the telling figure. The telling figure is miles driven, and from that the number of people, especially the particular people and the amount they are willing to dedicate to FSD.

My point is that they could not have added so many people in so many places with such easy to achieve without knowing about FSD Safety Scores, driving so many individually cumulative miles, in so many different traffic scenarios, and without having more accidents…

UNLESS FSD has dramatically improved in capability and safety.

You just couldn’t. You could not get THAT 100,000 to keep logging TONS of miles week after week without major show-stopping incidents if it was as bad as 10.2. Let alone pre-10.2! Which was the context of the original point showing the graph, that development has been FLYING since the architectural moonshot they took in the blank page rewrite.

Not only would they not want to, but the number of dice-rolls on weird traffic situations for that much less capable system would add up.

We all know the system will try to kill us some times. How often it tries and how many cumulative chances the system is given (across all the testers combined) matters to number of crashes. So does of course the vigilance and safety score of the monitor.
 
Last edited:
One question I have was at a certain point the car would not exceed 60MPH in the construction zone. Traffic was flying at 65-70MPH. I was set for 70. The car would only do 60, no more. Anyone seen this?? once it counts 500 cones it restricts itself?? First time i ever had a speed restriction like the old AP on back roads.
I got this on NOA on an Interstate loop. It’s a major project that has been running for a while. It slowed to 60 and wouldn’t let me change it. Oddly, I believe the construction speed limit is even posted higher than that but I’d have to circle back through to confirm. Even more strange, it only does this on the outer (counterclockwise) part of the loop. When I’m traveling on the inner (clockwise) part of the loop it doesn’t cycle me down to 60.
 
  • Like
Reactions: Reddy Kilowatt
Telsa is a nearly $1T company. Honda is $45B, GM is $60, Ford is $62B, and Totoya is $230B. How much more profitable do they need to be? They're making cars as fast as they can, and doing everything they can to get around global supply chain problems, while also totally revolutionizing the battery market. Other manufacturers have stopped production of entire lines because they can't get parts/materials, focusing on their most profitable models.
It's 1 trillion because of the stock market. He only became profitable in the last year or so. I don't have the patience to do the search but he EXPLICITLY said he's banking on FSD and robotaxis to make the company successful. He could also put some effort into more service centers and awful availability for service dates
 
  • Funny
Reactions: AlanSubie4Life
This gets mildly exciting when the car turns from a 50 mph express to 35 mph street and it thinks it should be going 50
Yeah, I've encountered two types of this unnecessarily fast acceleration with 10.69. One is like your situation of it keeping the speed limit when turning to a street with much lower limits, e.g., main street to residential street, so I know to rapidly scroll down the set limit during the turn. And what's even more silly is that it drops from say 40mph to 15mph to even make the turn only to speed back up trying to get to 40mph then slow back down to 18mph for a speed bump in this particular case for me.

Another situation more noticeable with 10.69 is it believes it needs to quickly reach the speed limit to match the speed of oncoming traffic after a turn except there isn't actually any traffic. No disengagement or intervention required for either of these, but this behavior would probably result in some eyerolls from the wife if she was in the car.

Even with 10.69, the only time I don't turn on FSD Beta is when I'm driving her car and she's awake. And she doesn't even have FSD Beta enabled on her profile. So in terms of minimizing jerk (changing acceleration -- especially switching between positive and negative), it seems like there's been some improvement to smoothness with 10.69, but not enough to be "spouse approved."
 
Does anyone have an idea why FSD 10.69.1.2 would suddenly become unavailable? No error messages, no diagnostics alert, auto steer, cruise control, nothing is available. No strikes showing up in the auto pilot panel. I've rebooted the car, and am at a loss here. I've driven it for ~30 minutes, no change.

Has anyone seen this?
I had to recalibrate my cameras to correct this.
 
Bingo. This is exactly my point. Apologies that I cast aspersions about your motives
I dunno, you might still not like my motives.

If I had to summarize though, I would say that I want people to be safe, and I do not want to see an accident caused by FSD Beta, or caused by someone’s misunderstanding of the capabilities. I want people to use both hands on the wheel and drive the car normally and avoid inconveniencing other road users when using FSD Beta. I think it is fine for it to be available to owners under those conditions. It’s been very interesting to see the progression, and will be interesting to see how the limitations will be resolved.
only one bollard boop.
I feel like this isn’t really completely accurate. There was also a small off-road excursion. (I feel quite likely to be legit.)


It’s either an inability to understand not everybody thinks exactly like they do, or a rejection of facts which go counter to a dearly held belief.

View attachment 853632
View attachment 853636
View attachment 853637
View attachment 853638

It would be nice to see data on miles per user per month for each version. Would also be nice to see the percentage of surface street driving where it is used. I would expect just a small increase over time so far.

Twitter is all very well, but a lot of stuff on there is not correct, just saying. People make claims. And even if truthful, a handful of self-selected people who do use it more will not have a huge impact on a pool of 100k.

That is why it is nice to see the data…which so far do not seem to suggest a large increase in usage with recent releases (but would be nice to normalize by user base, since that would make it easier to tell).
 
  • Like
Reactions: Dewg
It would be nice to see data on miles per user per month for each version. Would also be nice to see the percentage of surface street driving where it is used. I would expect just a small increase over time so far.

Since 10.2, I would use beta a lot to begin with on all my normal drives, within 1-2 weeks I’d stop using it and repeat for new versions. I’d argue I used it more early on because we were getting more frequent updates. With 10.12.2 for instance I was not a fan of the jerkiness and i had a good idea of it’s capabilities so after first couple weeks, I was bored of testing it out
 
  • Like
Reactions: sleepydoc