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.
Tried this sorry mess of a release again. It failed 100% of attempted left turns and right turns and botched stopping distance for one stop sign. So basically still pre-alpha level software where work hasn’t really begun in earnest for the car to make its own turns.
I'm new to FSDb, and I've been trying out 10.69.3.1 for the last few days. I've found it to be pretty inconsistent. I've had absolutely horrible drives, and then today I had several nearly perfect ones.

Overall, though, I'm enjoying seeing the car figure situations out (or not), and I'm definitely on guard to steer away from trouble or hit the brakes when needed (which I've had to do a few times already). It makes me pretty nervous how close I come to curbs on some turns, but so far so good. I can definitely envision FSDb continuing to learn quickly, and I'm hopeful V11 will bring more refinement and consistency.
 
Now that the report button is gone, how is Tesla going to know that FSD screws up here?
When the driver disengages FSDb using the steering wheel or brakes, the car can report this as a possible failure.

Tesla can also have the car report back even when driving manually if the planner would have done something differently than what the driver did.

And, I would assume that Tesla can task cars to report when they encounter specific situations. Say specific intersections, types of turns, etc. Easy to set up a tasking system with many layers of criteria. None require user input.
 
  • Like
Reactions: FSDtester#1
Tried this sorry mess of a release again. It failed 100% of attempted left turns and right turns and botched stopping distance for one stop sign. So basically still pre-alpha level software where work hasn’t really begun in earnest for the car to make its own turns.
I find it highly unusual and questionable how some reports are Soooo much different or exaggerated. Not saying you don’t have an issue just odd that like others I have driven probably 10k miles using FSDb on most every drive with few if any major issues. Not perfect of coarse but Far from 100% fail as some like to describe. Have most every street variable near me with exception for roundabouts but most everything else. Always wonder if it’s personal expectation issue or possibly a group setting configuration or other. Again not AlanSubi perfect but dam good for most all logical driving.
 
I find it highly unusual and questionable how some reports are Soooo much different or exaggerated. Not saying you don’t have an issue just odd that like others I have driven probably 10k miles using FSDb on most every drive with few if any major issues. Not perfect of coarse but Far from 100% fail as some like to describe. Have most every street variable near me with exception for roundabouts but most everything else. Always wonder if it’s personal expectation issue or possibly a group setting configuration or other. Again not AlanSubi perfect but dam good for most all logical driving.
Honestly if you are going in a straight line, it is mostly OK. Anecdotes of good performance are a combination of mostly straight drives with maybe some intentional non counting of manual disengagements for whatever reason. The most optimistic anecdotes (I.e., Whole Mars Catalog) are essentially just lies to curry favor with Elon.

Frankly I’m shocked that this level of non capability is considered sufficient for public testing. It needs far more lab only work to get to a level that would be considered remedial capability and then some time after that you graduate to beta to round out the rough edges.

Aside from all of that, clearly the software is clearly overfit to a number of CA and TX specific use cases that bare no resemblance to east coast road design.
 
I find it highly unusual and questionable how some reports are Soooo much different or exaggerated. Not saying you don’t have an issue just odd that like others I have driven probably 10k miles using FSDb on most every drive with few if any major issues. Not perfect of coarse but Far from 100% fail as some like to describe. Have most every street variable near me with exception for roundabouts but most everything else. Always wonder if it’s personal expectation issue or possibly a group setting configuration or other. Again not AlanSubi perfect but dam good for most all logical driving.
Got it the other day. Tried it for 90 seconds today. Will not turn it on again until the next major release and then only with no one else in the car, with zero traffic around me, and in very controlled conditions.

It is currently not fit for purpose and has yet to reach a suitable beta level of quality.

Tesla should be embarrassed that they expanded this out to more people.

New beta tester as of this week, using 2022.20.17. Quick summary, grade is F minus.

Tried it briefly, but it so thoroughly tried to kill me and wreak havoc across multiple straight-forward, non edge case situations that it isn't something I'll be using again until the next major version release.

Issues:
  • Left / Right turns are either too wide (go into a ditch) or to narrow (say hi to oncoming traffic)
  • Random freezes mid turn (say hi to being rear ended)
  • Maniac acceleration to traffic light, will it / won't it stop? Hard breaking insanity.
  • Navigate attempt through one-way intersection the wrong way
  • Not guaranteed to stay within the lines when / where it obviously should

General feelings of FSD otherwise (i.e., on high-way):
  • Impressively poor understanding of when to change lanes
  • Impressively indecisive follow through on executing lane changes at speed (very, very dangerous)
  • 50/50 chance it can appropriately handle merging in traffic
  • Unable to navigate high volume traffic situations when approaching backed up exit lanes
I think in general, FSD is probably still more an 'alpha' state where Tesla is just throwing things at the wall to see what sticks, and if it does stick, band-aiding over it repeatedly until it fails, and then we regress for a while and try again. If I had to guess, I'd say that Tesla's image recognition tech is probably not 'best in class', and that is probably more to do with the onboard hardware rather than anything else. ML isn't magic, just brute-force regression against an error minimization term, so garbage in, garbage out would apply.
I agree @Yelobird , see above from one of our new forum members. Notice days ago he said he would never turn it on again... but just today had more complaints from I guess more driving on beta he said he wouldn't do.
There are more negative posts from him, but I will spare you all the time.
I am not a moderator here, but have always enjoyed the positive and negative criticism. What I see above appears to be trolling. We don't like trolls here.
We already have 2 or 3, whom most of us have ignored a long time ago.

Advice : Don't be a troll, you don't know anymore than any of us using the system.
Some have been using it through multiple versions, and have seen the improvements and regressions. Don't be "absolute" in your opinions.
Nothing wrong with subjective anecdotes.
Also, don't start a fight with @AlanSubie4Life , @EVNow ... you will lose,
Or tangle with @Ramphex or @PACEMD They will destroy you with humor.
 
Last edited:
Honestly if you are going in a straight line, it is mostly OK. Anecdotes of good performance are a combination of mostly straight drives with maybe some intentional non counting of manual disengagements for whatever reason. The most optimistic anecdotes (I.e., Whole Mars Catalog) are essentially just lies to curry favor with Elon.

Frankly I’m shocked that this level of non capability is considered sufficient for public testing. It needs far more lab only work to get to a level that would be considered remedial capability and then some time after that you graduate to beta to round out the rough edges.

Aside from all of that, clearly the software is clearly overfit to a number of CA and TX specific use cases that bare no resemblance to east coast road design.
Ok to clarify I use it on ALL streets, turns, everywhere! Also in IL if that matters not CA. When I say not perfect yes on occasion I will occasionally give the accelerator a light bump to tell the software “yes, I confirm it’s good to go” but to me that is a minor confirmation (which is why we are there) vs a failure. My wife actually laughs because as soon as we back out of the garage I setup the GPS even to go to Target 3 blocks away. I want to use the software and Do everywhere.
 
I agree @Yelobird , see above from one of our new forum members. Notice days ago he said he would never turn it on again... but just today had more complaints.
There are more negative posts from him, but I will spare you all the time.
I am not a moderator here, but have always enjoyed the positive and negative criticism. What I see above appears to be trolling. We don't like trolls here.
We already have 2 or 3, whom most of us have ignored a long time ago.

Advice : Don't be a troll, you don't know anymore than any of us using the system.
Some have been using it through multiple versions, and have seen the improvements and regressions. Don't be "absolute" in your opinions.
Nothing wrong with subjective anecdotes.
Also, don't start a fight with @AlanSubie4Life , @EVNow ... you will lose,
Or tangle with @Ramphex or @PACEMD They will destroy you with humor.
I’m not a troll and it isn’t your place to try to moderate me. I’m providing my experience and frustrations with something that Tesla said was completely finished in 2016 and then every single year since then.

I have a background in machine learning and know more about it than your typical poster on here about programming and ML.

The bigger problem is that too many are unable or unwilling to hear any criticism whatsoever about Elon, Tesla, or FSD.

(moderator edit) start thinking critically about why some experiences are so divergent. There are ample examples about things going wrong, and there examples of at least some things going pretty well. But we know that Tesla had intentionally doctored video footage before and attacks all those who don’t tote the party line.

At the end of the day, you don’t have have to agree with me, but just because you don’t doesn’t mean I’m a troll.
 
Last edited by a moderator:
I also have been using FSDb for every drive since I got it last week. I live on an acreage with a 400-ft long paved driveway that ends at a left/right 45 mph road, and I've even tried FSDb while I'm still on my driveway, and it did just fine (stopped before the road, creeped forward, and made a good turn).

My most common issues so far have been bad planning decisions. My car sometimes inexplicably gets into a turning lane and actually completes a turn a half mile from where it was supposed to turn. I've been letting it do this, mainly to see how it recovers from this incorrect turn. This behavior is not actually dangerous, just inconvenient when I'm trying to get somewhere on a timeline.

I've also had the extremely jerky wheel and indecisive commitment on some left/right turns from stop signs. This has resulted in some disengagements as I try to keep my hands loosely on the yoke, and sometimes it hits my hand fairly hard and disengages when it's jerking around.

None of this has convinced me to give up on FSDb, and I'm still really having fun seeing how this technology works.
 
I’m not a troll and it isn’t your place to try to moderate me. I’m providing my experience and frustrations with something that Tesla said was completely finished in 2016 and then every single year since then.

I have a background in machine learning and know more about it than your typical poster on here about programming and ML.

The bigger problem is that too many are unable or unwilling to hear any criticism whatsoever about Elon, Tesla, or FSD.

Take the fanboy blinders off and start thinking critically about why some experiences are so divergent. There are ample examples about things going wrong, and there examples of at least some things going pretty well. But we know that Tesla had intentionally doctored video footage before and attacks all those who don’t tote the party line.

At the end of the day, you don’t have have to agree with me, but just because you don’t doesn’t mean I’m a troll.
I will refer you back to my post (moderator edit)

Another:
"But we know that Tesla had intentionally doctored video footage before and attacks all those who don’t tote the party line."

Source?

(moderator edit)
 
Last edited by a moderator:
I’m not a troll and it isn’t your place to try to moderate me. I’m providing my experience and frustrations with something that Tesla said was completely finished in 2016 and then every single year since then.

I have a background in machine learning and know more about it than your typical poster on here about programming and ML.

The bigger problem is that too many are unable or unwilling to hear any criticism whatsoever about Elon, Tesla, or FSD.

Take the fanboy blinders off and start thinking critically about why some experiences are so divergent. There are ample examples about things going wrong, and there examples of at least some things going pretty well. But we know that Tesla had intentionally doctored video footage before and attacks all those who don’t tote the party line.

At the end of the day, you don’t have have to agree with me, but just because you don’t doesn’t mean I’m a troll.
This is due is that you are such an outlier. All the videos on YouTube are pretty consistent and you can gauge pretty much the capabilities and incapabilities of the system. Please film your experiences and show it to us, else ask Tesla to check the set-up in your car. Your experiences are not good.
 
  • Like
Reactions: Sharps97
a few questions for anyone who knows please? I just received FSD 36.20 a few days ago.......

- Can I expect V11 as the next iteration of FSD beta?

- I'd read somewhere that Elon mentioned that V11 doesnt require users to touch the controls. Does this mean that touching the wheel wont be necessary?

Thanks in advance
 
  • Funny
Reactions: AlanSubie4Life
a few questions for anyone who knows please? I just received FSD 36.20 a few days ago.......

- Can I expect V11 as the next iteration of FSD beta?

- I'd read somewhere that Elon mentioned that V11 doesnt require users to touch the controls. Does this mean that touching the wheel wont be necessary?

Thanks in advance
Yes, you should get V11 along with the rest of us when it is widely released. Elon did say that this version will be hands-free, but many are skeptical that Tesla will allow that with wide release of this still somewhat buggy software (unless V11 resolves a LOT of bugs).
 
  • Like
Reactions: FSDtester#1
a few questions for anyone who knows please? I just received FSD 36.20 a few days ago.......

- Can I expect V11 as the next iteration of FSD beta?

- I'd read somewhere that Elon mentioned that V11 doesnt require users to touch the controls. Does this mean that touching the wheel wont be necessary?
- Can I expect V11 as the next iteration of FSD beta?

I believe the general consensus is V11 is the next release.
I mostly agree because it's allegedly on the .40 branch which would let all
The people with Fsd to get the update. Currently we are on .36 which means
People on the .40 won't get it until beta releases a .40 update.



- I'd read somewhere that Elon mentioned that V11 doesnt require users to touch the controls. Does this mean that touching the wheel wont be necessary?

Geez, I think most of us hope so. I did read a couple of weeks ago that an internal tester was using V11 with no nags, but I can't remember where I read it.
I do think with their most recent crackdowns on wheel weights and other cheat devices that it will be while before the nags stop.