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

FSD Beta 10.13

This site may earn commission on affiliate links.
And back to true 10.13 discussion since all the driving talk MUST be about 10.12 experiences and not 10.13.😉 Now that it has been sooooooooo long since the initial SINGLE (1) employee version showed up on TesaFi and was never officially released what are the odds of 10.1 being updated from week 16 to 20 firmware? Especially since 24 is now about to drop so on 20 most of the bugs should be ironed out by now.

Hell at the rate it is coming and needing just one more week every single week 10.13 could probably be updated to week 44 firmware version.:eek::oops:
 
Last edited:
I realized I sounded kinda whiny in my last few posts and wanted to clarify something. I love FSD beta. I use it practically every day. Unlike others I find it pretty reliable to more or less get me from point A to point B. And I don't even live in the Bay Area. I think it is so much better than it was when I first got it. There's just some inexplicable things that it does that make you feel confused or scared at times because it is not a human-like maneuver so it comes out as unexpected.

I can't wait to see what the next update does for the system but I'm hungry to get it right away.
 
If the car speeds up, then it's speed was being held back from the current maximum by traffic. So, speeding up, assuming that the new lane has the clearance, is an appropriate response.
That is exactly what is happening. Max speed is set at 40 MPH for example and speed of traffic is say 30-35 MPH. But still most people first make a lane change keeping a similar speed as they are doing and then accelerate when they are stable. This is through years of observation on the road. Accelerating while making the lane change almost feels like I'm going to scrape the bumper of the guy in front of me sometimes.

I say that this is odd because AP on highways does NOT do this. It makes a very smooth change of lanes at approximately the speed that you were going before and then accelerates. Sometimes I have the opposite problem with Navigate on Autopilot where it accelerates too slowly after the lane change is completed. But I hardly ever feel that Navigate on Autopilot makes an unsafe lane change and maneuver. I personally think this is the better way of doing things. It will be interesting to see what behaviors are adopted once we have single stack - the holy grail of v11.
 
  • Informative
Reactions: pilotSteve
And back to true 10.13 discussion since all the driving talk MUST be about 10.12 experiences and not 10.13.😉 Now that it has been sooooooooo long since the initial SINGLE (1) employee version showed up on TesaFi and was never officially released what are the odds of 10.1 being updated from week 16 to 20 firmware? Especially since 24 is now about to drop so on 20 most of the bugs should be ironed out by now.

Hell at the rate it is coming and needing just one more week every single week 10.13 could probably be updated to week 44 firmware version.:eek::oops:
26 pages of discussion without anyone having 10.13. When 10.14 comes along, the discussion thread that is, perhaps it should be titled, "anticipating 10.14", followed by another thread called ?10.14 experiences" once the first non-Tesla employee installs 10.14?
 
26 pages of discussion without anyone having 10.13. When 10.14 comes along, the discussion thread that is, perhaps it should be titled, "anticipating 10.14", followed by another thread called ?10.14 experiences" once the first non-Tesla employee installs 10.14?

I was annoyed when this thread was created just shortly after 10.12 was released. It gives people yet another thread to post off-topic stuff. Maybe when 10.13 actually drops, someone should create a "Real 10.13 Thread".
 
  • Like
Reactions: VanFriscia and tmoz
And back to true 10.13 discussion since all the driving talk MUST be about 10.12 experiences and not 10.13.😉 Now that it has been sooooooooo long since the initial SINGLE (1) employee version showed up on TesaFi and was never officially released what are the odds of 10.1 being updated from week 16 to 20 firmware? Especially since 24 is now about to drop so on 20 most of the bugs should be ironed out by now.

Hell at the rate it is coming and needing just one more week every single week 10.13 could probably be updated to week 44 firmware version.:eek::oops:
Right This ☝🏻

 
I like it this way. It’s fine. Being off topic is not limited to this thread. Where else would one post about 10.13 and efforts to improve unprotected lefts? Totally on topic and may as well have people’s thoughts on it prior to release in the same place as their thoughts post release. Of course there is off topic stuff, that is normal.

We are able to navigate threads. We can do it.

Tying my last off-topic post back to 10.13, I expect that 10.13 will also be jerky in every aspect of driving, just like 10.12. I expect only marginal improvements at best. My hope is that this is such an easy problem to fix that they are not bothering, since it increases safety levels and discourages use (especially with passengers) by keeping it jerky. Part of their strategy of maximizing safety and minimizing exposure.
 
Last edited:
I like it this way. It’s fine. Being off topic is not limited to this thread. Where else would one post about 10.13 and efforts to improve unprotected lefts? Totally on topic and may as well have people’s thoughts on it prior to release in the same place as their thoughts post release. Of course there is off topic stuff, that is normal.

We are able to navigate threads. We can do it.

If you look at the OP, that post should have gone into Elon beta tweets. Discussion about Chuck's ULTs could have gone in a thread called Improving ULTs. Not saying current thread can't work; it just sucks for people who aren't regulars who just want info about how 10.13 actually performs. Agree that off-topicness is everywhere. It's actually kind of a clusterf*k. I used to moderate androidforums.com. Keeping things tidy is definitely possible. It just seems like the moderation culture here is very hands-off. Which I'm ok with from a policing perspective, but I think TMC could do better with organizing content in more real-time.
 
  • Like
Reactions: AlanSubie4Life
The only thing smooth I’ve seen the car do in city driving is accelerating from red lights. Not nearly fast enough for me. But everything else is jarring.

Edit: I hope 13 fixes this as well as lane choices. There I tied it back to the thread. lol

It's gotten better. The first safety score cohort release (10.2) was extremely jerky.

my car has always done well decelerating for a red light. stop sign not so much. Lately it's doing a lot better modulating speed around curves.
 
Edit: I hope 13 fixes this as well as lane choices. There I tied it back to the thread. lol
I doubt it will fix that red light acceleration but hope I am wrong since it’s a major reason for accelerator use.
It's gotten better. The first safety score cohort release (10.2) was extremely jerky.
Yes.

While it’s definitely a lot smoother, I feel like it has largely plateaued now, which is why I expect minimal improvement in 10.13. (Trying to stay on topic too, lol.)

And when I say it is jerky at everything, I don’t mean it is not sometimes smooth, I just think that every class and subclass of maneuver has significant rates of failure in this metric - instances where it jerks the user around (or at least is less smooth than a competent human - and it should be smoother than the best human).
 
  • Like
Reactions: nvx1977
I doubt it will fix that red light acceleration but hope I am wrong since it’s a major reason for accelerator use.

Yes.

While it’s definitely a lot smoother, I feel like it has largely plateaued now, which is why I expect minimal improvement in 10.13. (Trying to stay on topic too, lol.)

And when I say it is jerky at everything, I don’t mean it is not sometimes smooth, I just think that every class and subclass of maneuver has significant rates of failure in this metric - instances where it jerks the user around (or at least is less smooth than a competent human - and it should be smoother than the best human).

some elements of FSDb discomfort seem like it'd be so easy to fix. Of course it's easy to armchair quarterback, but even Musk has said that comfort is currently not a priority when the car is still making basic wrong decisions. I guess that makes sense, but hard to imagine why these things can't be worked on in parallel.
 
but hard to imagine why these things can't be worked on in parallel.
It’s presumably just a flag in the software they can turn on and off. (Elon runs with it on I am sure!) As you say, super simple relative to other challenges.

Tesla has a robust commitment to safety and want to be sure that only committed testers are using this pre-release beta of FSD Beta. So this enables that goal to be achieved.

Normal users will turn it on, say “WTF?”, and then promptly stop using it.
 
I have one in my neighborhood that is worse. Big bushes right out to the street. Only way that will ever get resolved is with a camera at the front of the car, say at the headlight aimed perpendicular to the car.. I proved this with a gopro mounted out in front. I have to pull out half a car length before I can see a car 80 feet coming at me. They usually lay on the horn when they see my hood poking into the lane of traffic.
These sort of intersections shouldn't exist. When there's high occlusion and fast traffic it's a recipe for disaster. My wife's car is an Audi A4 Allroad and it sits you further back than where my side cameras are on my car. I got in a situation in a town I hadn't visited before in her car, where traffic was 55mph and there was simply no safe zone to pull out and see cross traffic. The people behind me were honking and I pulled out and got honked at by cross traffic and I still couldn't see if it was clear to go because of a high wall and bushes on either side. Navigation told me to make a left and I had to make a right to even get out of there. When driving through an intersection feels more like Russian roulette we need to rethink these.
 
Tying my last off-topic post back to 10.13, I expect that 10.13 will also be jerky in every aspect of driving, just like 10.12. I expect only marginal improvements at best. My hope is that this is such an easy problem to fix that they are not bothering, since it increases safety levels and discourages use (especially with passengers) by keeping it jerky. Part of their strategy of maximizing safety and minimizing exposure.
They have pretty much stated they are striving to get the car to handle all the various driving situations before they smooth out the rough edges. But yes, oddly (and though not deliberately) by keeping it jerky they do tend to keep the test drivers alert :)

I had FSD on a long (170 mile) drive last week .. the roads were not complex, mostly rural winding roads with good visibility, and the car drove impeccably ..so much so I realized quite a few times I was really not paying as much attention as I should (and normally do).

Perhaps even after FSD is released Tesla need to make the car occasionally do something a bit dumb (though not dangerous) to keep the driver alert? Or play C&W music if you dont pay attention enough???
 
It's gotten better. The first safety score cohort release (10.2) was extremely jerky.

my car has always done well decelerating for a red light. stop sign not so much. Lately it's doing a lot better modulating speed around curves.
My biggest complaint with acceleration is how jerky in neighborhoods it is where turns are 1 or 2 tenths of mile apart. When I leave my driveway I don't even enable FSD until I leave the neighborhood otherwise FSD is trying to accelerate to 30mph then having to quickly slown down to make turns that are a 10th of a mile apart. A bit like a roller coaster:)
 
Looks like the "bag of points" architecture has hit a ceiling as well. Now well need to wait for another redesign before next level of improvements :rolleyes:

Pretty sure we have hit a ceiling either way with HW3 and camera setup. I dont see them solving NYC FSD (ie 10+ cars + 30+ people and bikes in a single scene the car has to navigate around) with this hardware.

Will probably take HW4 + Dojo fully running + new cameras to get to a point of NYC robo taxi. Until then they'll probably just perfect the easy suburban routes everyone posts on YouTube
 
  • Disagree
  • Like
Reactions: aronth5 and edseloh