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.
The thing is, lower jerk is objectively better. It's not really subjective. And given FSD's sometimes early slowdowns, it's actually possible for it to make very low jerk stops in many scenarios. But it doesn't. In fact, it does very suboptimal jerk profiles (you can tell from how quickly the regen bar changes - you want to see smooth monotonic increases in that bar length, followed by monotonic decreases). It shouldn't be bouncing around (it does).
Actually yes, I agree with that statement so I guess I was not clear with my point. I'll just give up. :)
 
Deep lane guidance has f***ed us. This was fine before that garbage. It no longer believes its lying eyes.
That's some serious forking going on there. Here's looking back showing the 4 lanes start together and before that it looks like a lane getting really wide:

1 into 4 fork.jpg


If it is related to deep lane guidance, I would think that module is incorrectly saying to ignore map data as this is a new regression. Additionally, 10.69 had some focus on forking lanes too, but maybe the training data there focused on 1-to-2 forks and not so much this 1-to-4?
 
  • Like
Reactions: sleepydoc
One more for tonight…..the “lane planning” has been a fiasco from day one through 10.69.2.

On my daily commute after exiting the expressway there is a light with 2 left turn lanes. My next turn is less than a couple of hundred feet and is a right turn. IF I allow FSD to select the lane at the exit light it always picks AND moves to the left hand lane and then sees that it has to immediately move back to the right lane for the upcoming right turn. Does it every time.
D881A40A-435E-40F0-84FD-DB4EEB1B19D8.jpeg
 
One more for tonight…..the “lane planning” has been a fiasco from day one through 10.69.2.

On my daily commute after exiting the expressway there is a light with 2 left turn lanes. My next turn is less than a couple of hundred feet and is a right turn. IF I allow FSD to select the lane at the exit light it always picks AND moves to the left hand lane and then sees that it has to immediately move back to the right lane for the upcoming right turn. Does it every time.
View attachment 852622
But yet…Elon says this is 🔥🔥🔥
 
  • Like
Reactions: boonedocks
Depends on your perspective, .69 is very impressive for being better than 10.12, meanwhile adding brand new architectures (that are necessary for future progress).

The accel / decel issues are simple fixes (along with most of my main issues). I'd hazard to guess that the AP team purposely makes a version less comfortable so that people avoid using it or pay more attention, lol.
 
Last edited:
10.69 is fine.

It is just not very good.

It is exactly what we expected. Lots of new capability, but lots of work still to do.

Give it a few years and it will be a lot better. It has improved quite a lot in a year, even though as far as utility goes it is no different than a year ago.

There is plenty of time; there is no current competition.

You have to appreciate it for what it is, not what you want it to be, then turn it off.
I just want it to be as what Elon has been telling it is and will be by the end of the year… end of every year for the last four years…
 
  • Like
Reactions: 2101Guy and Ramphex
Speaking of vision and navigation not working together, I thought this was funny today. Blue line was the nav route, green line was the line the beta took after getting confused about the left-hand turn. Where the 2 lines meet is a stop sign, which the beta respected correctly. Meanwhile the nav did not notice and kept talking about making a right turn.

Screen Shot 2022-09-14 at 6.42.13 PM.png
 
  • Informative
Reactions: willow_hiller
I wish they would update the cameras. My $20 security cameras have amazing nighttime imagaing with near no light. I actually keep the IR off and I get near daytime quality in full color.

I just want it to be as what Elon has been telling it is and will be by the end of the year… end of every year for the last four years…
😂
 
  • Funny
Reactions: jebinc
You are stuck in the Stone Age! 50mph in 8 seconds is completely normal now. In 6 seconds, slightly peppy.
👏Depends on where you are👏

I want you to have that as an option. Tesla needs to make it happen, no doubt about it.

But some of us need a slow option because of where we are. I used to routinely smoke people in my Prius where I’m at. Keeping a safety score of 100 around here was literally no effort at all. I definitely don’t want FSD to go any faster than it already does.
 
Two great 10.69.2 drives this evening. The first had no disconnects, but did have an application of accelerator where it hesitated too much trying to make a quick two lane shift to the left to enter a tollway. Later, after exiting the tollway, FSD picked its way successfully through an intersection that was partially blocked due to stopped traffic and found an open lane to turn left into. 10.12.2 would've had no chance here. Then, at the end of the drive, FSD made a brilliant forward left turn at a flashing red despite their being serious occlusion from a car wanting to do the same thing from the opposite direction. FSD waited until the oncoming lanes were clear and made the turn safely. Another maneuver that 10.12.2 would have failed miserably at.

Coming home in the dark, I did have one disconnect at a left turn with two left turn lanes. FSD was turning into a lane I did not expect and I was afraid it might cut another car off so I took control before I saw that there was no threat. My bad, but this was new behavior at a familiar location and it startled me. I need to drive this intersection a few more times to get confidence as I'm still not sure it's doing the 'right thing' here. Other than that, no interventions of any kind.

Total was about 15-20 miles on FSD beta. Very pleased with its performance
 
Post some videos. I’d love to see a car consistently stopping smoothly, without the behavior I describe. As I said, sometimes it does stop smoothly (rough guess is 20-30% of the time; higher with lead vehicles). So a representative sample would be good. Not with lead vehicles. So far every video of reasonable length I have watched shows the behavior.
Sorry, I don’t have any good way to make a video to post.
Me too! That’s why it is so frustrating. So you’re saying you are not seeing FSD Beta using the brakes? I often (not always!) see it do an initial pulse of strong regen, but not maxed out, then dial it back, then bring it back in with brakes as well. (All easy to see now; before I had to guess.) I’m typically talking about slowdowns from 50mph which is the typical surface street speed around here. 20mph stops tend to be handled better (but are slow).
No, I see it use the brakes because it brakes too late and needs to augment regenerative braking. I‘d prefer it brake earlier.
I like to drive assertively and never use the brakes.
virtually impossible to do for a human much less a computer.
There’s a huge difference between 0-50 in 10-15 seconds and 0-50 in 6-8 seconds. The latter can be extremely smooth and comfortable and is definitely not aggressive.
0-50 in 6-8 seconds is definitely aggressive. That’s close to or beyond maximum acceleration for many cars on the market. Just because Teslas can accelerate in half that time doesn’t mean it’s not aggressive.

The problem is that driving is not a science, but more a personal experience, and highly subjective. … What is slow acceleration to one person is "just right" for others, and even "aggressive" for others.

The rest is subjective. How you reach the speed limit is entirely up to you. Some enjoy gunning it and getting to the limit as fast as possible. Some enjoy moderate acceleration, and others prefer a more leisurely ramp up.

I view FSD Beta from the standpoint of capability. Does it follow the law, and do so safely? The subjective items I try to dismiss as if I was in an Uber and someone else was driving. So, it didn't accelerate as fast as I normally would. So, it took that corner a little faster than I would normally. Did it get me to where I wanted to go? Did it do it safely? That's all that really matters to me. But that's my personal experience.
YES
haha. My first car, an Acura Integra, did 0-60 in 9 seconds and that required skill to achieve. 6-8 seconds 0-50 is aggressive (not that I don't do it every time I can :D)
My first car was 1980 Toyota Corolla. After that I had a ’95 Escort wagon. In both of those cars you were lucky to do 0-60 in less than 15 seconds.
You are stuck in the Stone Age! 50mph in 8 seconds is completely normal now. In 6 seconds, slightly peppy.
Hardly. You’re biased by driving a Tesla. Go out and drive a Toyota Camry (without flooring it)
Objectively, I would like to see FSD Beta not use the brakes when stopping in most cases (obviously there are scenarios with lights turning yellow or on downhills where it may be physically impossible). Can we agree that that would be reasonable?
Yes, that would be reasonable, but I suspect you’d be complaining about how conservative the braking is.
 
👏Depends on where you are👏

I want you to have that as an option. Tesla needs to make it happen, no doubt about it.

But some of us need a slow option because of where we are. I used to routinely smoke people in my Prius where I’m at. Keeping a safety score of 100 around here was literally no effort at all. I definitely don’t want FSD to go any faster than it already does.
I couldn’t decide whether to agree or to laugh at the image of you smoking people in your prius! (It wasn’t a toddler on a trike, was it?)
 
  • Funny
Reactions: FSDtester#1
virtually impossible to do for a human much less a computer.

Not sure what you’re talking about. I do it all the time. See the video above. That’s extremely aggressive driving by your standards and I did not touch the brake once. Standard practice. Tesla makes it easy.

Go out and drive a Toyota Camry (without flooring it)
I have a Toyota Highlander that weighs something like 5000 pounds and I have no problem getting to 50mph in 7 seconds.

No, I see it use the brakes because it brakes too late and needs to augment regenerative braking. I‘d prefer it brake earlier.
That’s sometimes what happens but not always what happens. I’ll post some videos at some point, or you can just watch the variety of videos already available.

Specifically: sometimes it uses regen, then inexplicably eases off of regen significantly for a second or so, then uses brakes and regen. That is clearly without any debate not optimal and results in excess brake use. It’s bizarre and I know it is not just me since it is in many of the videos I see.
 
Last edited:
That's some serious forking going on there. Here's looking back showing the 4 lanes start together and before that it looks like a lane getting really wide:

View attachment 852621

If it is related to deep lane guidance, I would think that module is incorrectly saying to ignore map data as this is a new regression. Additionally, 10.69 had some focus on forking lanes too, but maybe the training data there focused on 1-to-2 forks and not so much this 1-to-4?
It made it today. There was no lead traffic though. I bet it would have failed if there had been.
 
Do we agree that the miles driven per FSD beta user per day has been close to constant?
I highly doubt that. You’ll have to provide some sort of rationale. It was white-knuckle chimpanzee with a steering wheel when it first went Beta. You’re saying people are using it no less often now that many of their drives are disengagement or even intervention free? No way.
I think you may be misreading the graph. It’s cumulative miles. So constant slope means constant miles per day. The inflection points are when Tesla released it to more users.

Not quite. Those aren’t inflection points. Those are data points when Tesla reported the number.

The overall shape of the curve is clear though, the miles driven are growing at an increasing pace.

No surprise there. Our knowledge of the context of those miles let’s us understand how it correlates to system ability.

I don’t understand your conclusion about capability on varied roads. FSD beta is not geofenced and there is no evidence that the original group of beta testers was chosen to restrict usage to certain areas and situations.

I said varied roads and traffic scenarios. You can drive through the same intersection in the same direction 15 times and encounter 15 different traffic scenarios. 15 times along a 1mi route and you can encounter dozens of different situations.

(Though I don’t think it’s a stretch to also guess that 100,000 beta testers have hit more roads than 1,000 did.)

But the sheer number of road and traffic situations (which includes other cars, other road users, the time of day, weather, changes to the road itself, etc) that have been encountered after all those miles has to be simply gobsmacking.

Not nearly enough to capture all the edge cases, but an incredible number nonetheless. And with only that one bollard boop.

It seems you’re arguing that FSD is no better now than it was at the start of the year. This is a strange theory that does not seem to be remotely evidence-based. I’m not going to try to convince you otherwise.
 
It seems you’re arguing that FSD is no better now than it was at the start of the year. This is a strange theory that does not seem to be remotely evidence-based. I’m not going to try to convince you otherwise.
It seems you’re arguing that FSD is much better now than it was at the start of the year. This is a strange theory that does not seem to be remotely evidence-based. I’m not going to try to convince you otherwise.