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.
It's been 6 days since Elon tweeted "next week" for 10.69.3, and he seems to be ignoring questions from folks like Chuck Cook about it.

Given we haven't seen a release to employees yet, I'm not too optimistic about it actually releasing in the next few days... But I'd love to be surprised.
Your dates are off by a LOT:

90EBF348-EC07-4A4D-94AC-6CA229C1C1FF.jpeg
 
  • Like
Reactions: El joe
Your dates are off by a LOT

He gave an update on October 21:
Anyway, another example of my failure posted earlier. Fun times. No danger here of course. Just cannot exit.

I encountered an odd bug today, and given your circumstances I wonder if it's affecting your turn as well.

On a stretch of highway that's after an on-ramp and right before the highway splits into express and local lanes, my car switched into NoA display, but was still highlighting cars in blue as if it was on FSDb. It displayed "Upcoming lane change" to the left, and then indicated and merged to the right. And then immediately merged back to the left. So I think both NoA and FSDb were active simultaneously and pulling the car in opposite directions.

So maybe FSDb is still running in the background while you're on NoA. FSDb probably hasn't been optimized to take exit ramps like NoA does, and maybe it's sending the conflicting instructions to your car to abort that exit at the last second.
 
I don’t know where this misinformation comes from. Obviously they can see much further than this otherwise there’s no hope. They need to be able to see at least 500 feet, and they pretty clearly can get close to that already, and preferably see 500-800 feet.
I'll gladly take your metrics - it means our cars' B-Pillars can see about 500 feet, which gives us 5.68 seconds. That should be enough.
 
That should be enough.
It should be, but if you observe their videos and look at crossing speed, you’ll find this is tight.

Basically for humans over 5-6 seconds or so starts to feel “comfortable”. So it is right on the edge, because FSD does not move nearly as aggressively as human drivers after making the go decision.

Has been discussed previously and can example Chuck’s videos and you’ll see these numbers are not made up. Have had previous discussion about increasing crossing speed by 20-30% or so (and I’ve done my own tests to confirm it is possible to do this and still use median).

Hopefully we’ll get that in 10.69.3; it is actually essential for high success rates - there is not another way to do it.
 
I don’t know where this misinformation comes from. Obviously they can see much further than this otherwise there’s no hope. They need to be able to see at least 500 feet, and they pretty clearly can get close to that already, and preferably see 500-800 feet.


No clearly not. They need 5-6 seconds at least.


Anyway, another example of my failure posted earlier. Fun times. No danger here of course. Just cannot exit. Reported of course.

Curious if reproducible with HW2.5 or other vehicles; many must have noticed this in SD already. It is a common exit for people with AP/NOA/FSD. Though it is very possible no one uses those features!

I've had the same and similar issues at this exit. Day and night.

Just recently when it was a severe exit lane departure, I let it play out under close supervision since there were no cars nearby, and I had time for the detour. It veered left out of the exit lane and skipped the exit.

It's consistently bad there, so I'm always ready and normally have to intervene. Other drivers who see this must think I'm really indecisive about taking the exit.

I've reported it many times for many months now.
 
I've had the same and similar issues at this exit. Day and night.

Just recently when it was a severe exit lane departure, I let it play out under close supervision since there were no cars nearby, and I had time for the detour. It veered left out of the exit lane and skipped the exit.

It's consistently bad there, so I'm always ready and normally have to intervene. Other drivers who see this must think I'm really indecisive about taking the exit.

I've reported it many times for many months now.
Thanks for the confirmation. Must be a map issue since the car really doesn’t seem to think there is an exit there on the visualization. I haven’t looked super carefully at the videos to see how it behaved, but looks messed up.

I don’t think it is the pavement color difference since that is very common and you say it happens at night too.

Bizarre. Not an FSD issue presumably (since using AP) but would be nice to hear from someone without Beta. Or HW2.5. I have only ever taken this exit on AP/NOA when running FSD Beta.

And I guess technically I have never taken the exit with AP/NOA, lol.

As usual Tesla ignores reports. I think they are a complete waste of time. If they needed to know where it was screwing up they could just figure it out themselves trivially; they don’t need us hammering a silly button. Zero purpose; it will always be obvious from telemetry.
 
Last edited:
He gave an update on October 21:


I encountered an odd bug today, and given your circumstances I wonder if it's affecting your turn as well.

On a stretch of highway that's after an on-ramp and right before the highway splits into express and local lanes, my car switched into NoA display, but was still highlighting cars in blue as if it was on FSDb. It displayed "Upcoming lane change" to the left, and then indicated and merged to the right. And then immediately merged back to the left. So I think both NoA and FSDb were active simultaneously and pulling the car in opposite directions.

So maybe FSDb is still running in the background while you're on NoA. FSDb probably hasn't been optimized to take exit ramps like NoA does, and maybe it's sending the conflicting instructions to your car to abort that exit at the last second.
New “next week” update coming any day now 😂🤣
 
As usual Tesla ignores reports. I think they are a complete waste of time. If they needed to know where it was screwing up they could just figure it out themselves trivially; they don’t need us hammering a silly button. Zero purpose; it will always be obvious from telemetry.
Bold claim verified!

Here is a little follow up speculating about what might happen instead:

 
Last edited:
Thanks for the confirmation. Must be a map issue since the car really doesn’t seem to think there is an exit there on the visualization. I haven’t looked super carefully at the videos to see how it behaved, but looks messed up.

I don’t think it is the pavement color difference since that is very common and you say it happens at night too.

Bizarre. Not an FSD issue presumably (since using AP) but would be nice to hear from someone without Beta. Or HW2.5. I have only ever taken this exit on AP/NOA when running FSD Beta.

And I guess technically I have never taken the exit with AP/NOA, lol.

As usual Tesla ignores reports. I think they are a complete waste of time. If they needed to know where it was screwing up they could just figure it out themselves trivially; they don’t need us hammering a silly button. Zero purpose; it will always be obvious from telemetry.
Alan,

A thought on your report:

I just watched your video again, and wonder if this is related to the AutoPilot swerve behavior reported repeatedly elsewhere for years (e.g. Annoying FSD/autopilot behavior with on-ramps). That is when driving in the right lane of a freeway pasting a merging on-ramp. When the right lane edge line ends at the start of the merge, the Tesla perceives it's lane has widened, so it drifts to the right to stay centered, into the merge area. But then it slides back left as it sees the lane narrow, and in the process freaks us out by swerving toward cars in the second lane....

In your crazy behavior at an exit, maybe the car is not aware that this is actually an exit lane and perceives instead a widening lane, and so centers itself in that lane. But when the car ahead unblocks the view of the beginning of the painted divider between the through and exit lanes, it finds itself headed straight for that divider and has to quickly decide which lane to take.

Could this be what's happening? If it is what is causing your situation, hopefully the eventual FSD based highway lane keeping will improve both on and off ramp lane keeping. We can hope...

My two cents on this: I think Telsa's slavish lane centering is not a great design anyway. The car should instead be choosing a line to drive which minimizes swerving and maximimizes turning radius, by allowing it's position within the lane width to vary. In both your case and the on ramp case, the smooth line would be straight ahead, staying with the left side lane marking. In the curvy road case, the driving line would enter and exit the curves to the outside edge of the lane, and getting close to the inside edge at the middle of the curve, to enlarge the turn radius and hence decrease the turning G forces. Like race car drivers do, I think. Slavishly trying to stay centered is a novice driver's technique which stems from focusing one's gaze too close in front of the car instead of as far down the road as possible, and letting peripheral vision do it's job of keeping one off the lane edges.

SW
 
But when the car ahead unblocks the view of the beginning of the painted divider between the through and exit lanes,
Doesn’t matter if there is lead traffic. Fails anyway, at least some of the time (never have had a success).

Anyway I don’t the lane centering is what is happening here. It’s taking the exit. Then it is not.

Guess have to look at what the blue tentacle does in the AP case with merging lane but I would guess not similar.
 
  • Like
Reactions: Oil Freedom