FSD Beta 10.69

[momo3605]

Not sure why but this turn seems to have regressed in 10.69. Basically the right turn is up ahead but FSD wants to dive into the bus lane so it’sa disengagement every time.

This was ironically ok on 10.12. I was hoping the road marking improvements would’ve helped in 69.3

 

[JulienW]

Not sure why but this turn seems to have regressed in 10.69. Basically the right turn is up ahead but FSD wants to dive into the bus lane so it’sa disengagement every time.

This was ironically ok on 10.12. I was hoping the road marking improvements would’ve helped in 69.3

Not sure how it is supposed to work but the bike/bus lanes seem to be painted wrong. They are striped which means they are shared and you can drive over and into them. They should be solid lines if you are not supposed to enter them.

 

[sleepydoc]

Not sure how it is supposed to work but the bike/bus lanes seem to be painted wrong. They are striped which means they are shared and you can drive over and into them. They should be solid lines if you are not supposed to enter them.

I don't think it's the bike lane that's the issue - it's the lane that the camera and car are in. It appears the lane is a right turn only lane for the entire block but it's marked a bit strangely so it looks like the road goes from two lanes to one.

 

[nvx1977]

I have a thought experiment for anyone that wants to give it a try. Especially those who've been on Beta for about a year now. Imagine your experience, from the moment you were invited into FSD Beta, if Elon didn't make any public statements at all. He's the man behind the curtain, still running things, but he just never opens his mouth, nor makes any statements at all. Updates still roll out as they did, with release notes as they are.

How would your FSD Beta experience change? Would your expectations be different? Would you be more excited for updates, or more impressed with capabilities? Would you be more disappointed as updates rolled out? Or would your feelings be the same regardless?

I've never bought into Elon's hyperbole, even back in 2016. I was always cautiously skeptical and still am. I think FSDb needs another 1.5 years to be a solid L2 system based on what I saw since Oct 11, 2021 when I got v10.2. I would expect in that timeframe FSDb will be able to understand a lot more signage, and the car will rely more on local road data, at least for known trouble areas where vision has little-to-no lead time before a decision needs to be made.

But 1.5 years for a solid L2 system would be amazing IMO, especially one that works practically anywhere. I don't live in city limits, so competitors that offer geofenced city robotaxis don't interest me at all. I also never plan to use my car as a robotaxi.

 

[aronth5]

I've never bought into Elon's hyperbole, even back in 2016. I was always cautiously skeptical and still am. I think FSDb needs another 1.5 years to be a solid L2 system based on what I saw since Oct 11, 2021 when I got v10.2. I would expect in that timeframe FSDb will be able to understand a lot more signage, and the car will rely more on local road data, at least for known trouble areas where vision has little-to-no lead time before a decision needs to be made.

But 1.5 years for a solid L2 system would be amazing IMO, especially one that works practically anywhere. I don't live in city limits, so competitors that offer geofenced city robotaxis don't interest me at all. I also never plan to use my car as a robotaxi.

Agree completely.

Just once though I'd like to get an FSD update and say "wow" lots of improvement. V11 may do that for the highway but I'm not expecting anything great for city/streets with V11.

 

[AlanSubie4Life]

Just once though I'd like to get an FSD update and say "wow" lots of improvement.

This is just not going to happen with release cadences of 1-3 months. We would need releases every 1-2 years.

V11 may do that for the highway

I expect some significant regressions on V11 highway. Of course it will do some things that AP never was able to (like the exit above, hopefully) but for overall functionality I expect to be degraded on initial release. And then eventually it will be better.

 

[jebinc]

Every expansion has been described as “disaster waiting to happen”. I hope doesn’t come true now.

Time will tell; if V11 gets released this year. If it is delayed until next year, that may be a blessing in disguise

 

[Mardak]

I expect some significant regressions on V11 highway… but for overall functionality I expect to be degraded on initial release

Indeed, I tried activating FSD Beta 10.69.3.1 on a divided highway at 60mph, and spouse immediately noticed on the first curve as it couldn't smoothly maintain its speed needing multiple jerky adjustments. Once we got to the interstate with 70mph speed limit, the same types of curves were handled fine by NoA.

 

[jebinc]

This is just not going to happen with release cadences of 1-3 months. We would need releases every 1-2 years.



I expect some significant regressions on V11 highway. Of course it will do some things that AP never was able to (like the exit above, hopefully) but for overall functionality I expect to be degraded on initial release. And then eventually it will be better.

Sorry to disagree with your first point. I oped out in October last year and just got back in, after more than a year hiatus. Sure, it is smoother, but not a large step function change from last year. Lest us not kid ourselves; this is still a struggling mightily L2- system.

 

[Rikster]

I think I've been influenced by Elon's confident tweets, more so originally before I started being active on this forum. I think that people who don't closely follow Tesla happenings or Elon tweets, they really don't understand what's happening in the world of FSD. If I show a friend how my car currently drives itself (mostly), he/she is normally pretty blown away.

As for me, I use FSD just like I used AP before it, as a driver aid and not as a fully autonomous, self-driving car. I've only had FSDb for a week or so, but I'm getting pretty familiar with which things it does consistently well and which things it is unpredictable (and dangerous) at. I simply find myself disengaging in advance when I know FSDb is going to screw up something delicate or complicated. For 90% of my driving, though, it's amazing! I'll give it another try with each new release, and hope that it gradually gets better at some of the challenging stuff so I can add those things to my normal drives.

There's no way I'll ever use my car as a Robo-Taxi, so FSD to me is just a really nice driver aid.

 

[AlanSubie4Life]

I oped out in October last year and just got back in

That’s only 1 year. Maybe I should have said 2-3 years. I was trying to be generous.

That being said I think there has been a lot of improvement since last year, particularly with turning (because it was so bad a year ago, not because it is good now). It still struggles mightily with stopping and going, and I think that is frustrating and obvious to the point that it makes the turning improvements less enjoyable.

I guess what I am saying is that it is possible that there has been a step change since a year ago, but it is still so woefully inadequate that most people would decide it is struggling mightily (and they would be correct in my opinion), and just turn it off. (This is what I do on city streets typically, unless no one is around - I am just too embarrassed otherwise.)

 

[nvx1977]

Sorry to disagree with your first point. I oped out in October last year and just got back in, after more than a year hiatus. Sure, it is smoother, but not a large step function change from last year. Lest us not kid ourselves; this is still a struggling mightily L2- system.

I actually think your return experience with FSDb is not typical. For most of us it's gotten a lot better. Around 10.10, lots of people started reporting 0-disengagement drives. That was unheard of when we got 10.2. And don't forget the 10.3 fiasco that caused AEB events for no reason and FSDb glitching out in the middle of drives.

 

[JulienW]

Damn I miss 69.2.4 a lot. If I knew then what I know now I would have never updated to 3.1. Also more importantly I would have picked the correct Lotto numbers.

 

[AndrewZ]

What has happened is exactly what Tesla said would happen: the Beta testing program was there to refine the user experience (through the intervention/disengagement and camera button data + revision) prior to a *final* release to the full fleet, but the company had no illusions about fundamental OEDR changes facilitating a shift of DDT ownership away from the driver.

 

[Dewg]

Damn I miss 69.2.4 a lot. If I knew then what I know now I would have never updated to 3.1. Also more importantly I would have picked the correct Lotto numbers.

Always wait for the .2 version.

 

[2101Guy]

I
There's no way I'll ever use my car as a Robo-Taxi,

That is correct

 

[aronth5]

TI expect some significant regressions on V11 highway. Of course it will do some things that AP never was able to (like the exit above, hopefully) but for overall functionality I expect to be degraded on initial release. And then eventually it will be better.

Damn I miss 69.2.4 a lot. If I knew then what I know now I would have never updated to 3.1. Also more importantly I would have picked the correct Lotto numbers.

Funny how FSD brings out differences between owners. For me 69.2.4 was terrible when I netted out improvements vs. regressions. 69.3.1 definitely better for me except for 1 intersection I've documented in earlier posts.

 

[Supcom]

Funny how FSD brings out differences between owners. For me 69.2.4 was terrible when I netted out improvements vs. regressions. 69.3.1 definitely better for me except for 1 intersection I've documented in earlier posts.

I found no improvement on trajectories through intersections and significant new issues with hesitancy to enter intersections, crawling through intersections and wanting to stop in the middle of intersections.

I've also seen it enter intersections when cross traffic is closer than it should - especially when it gets in front of that traffic and slows way down instead of speeding way up. Maybe there's a sign error calling the requested_acceleration function!

 

[Gogoplata]

I found no improvement on trajectories through intersections and significant new issues with hesitancy to enter intersections, crawling through intersections and wanting to stop in the middle of intersections.

I've also seen it enter intersections when cross traffic is closer than it should - especially when it gets in front of that traffic and slows way down instead of speeding way up. Maybe there's a sign error calling the requested_acceleration function!

I've experienced exactly the same. I'm new to beta. Did previous iterations have these same issues?

 

[Supcom]

I've experienced exactly the same. I'm new to beta. Did previous iterations have these same issues?

10.69.2.4 was much better in this area. Hopefully, the next version will improve. It should be no surprise that new versions may regress or introduce new issues. Tesla is still making major changes to the system.