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.
First long (330 miles) drive with latest FSDb. It was great! It was terrible!

It is a great TACC + lane keep system, and it can change lanes with the press of a button on the yoke. Big roads, little roads, windy roads, hilly roads, it would slow down to a sensible speed on curves if needed, and stayed in its lane. Great; I was gruntled.

That said, I don't think it went correctly from one road to the next even once. Once it tried to switch to another road, even though the nav said otherwise. Another time it started to make a left turn, but then suddenly turned the wheel back to the right so hard that my shoulder hurt for about 30 minutes, and stopped in the middle of the road ( a red squirrel maneuver). I was disgruntled.

Driving through one small town it did a terrible job, but did better than the car in front of me, and about as well as I did the last time I drove through there. Okay, call it a draw

Definitely made the drive easier. Maybe next version will be fantastic?
 
It may not be legal but in reality, nobody does it. Here, I see people leaving their blinkers on forever, turning without a signal, signal and then turning in opposite directions; cops rarely pay attention to that unless they are speeding by a lot. ;)
But like with rolling stops NHTSA will make Tesla make FSD comply with the laws. So it will drive differently than most people. Which could result in it being unable to make some maneuvers.
 
But like with rolling stops NHTSA will make Tesla make FSD comply with the laws. So it will drive differently than most people. Which could result in it being unable to make some maneuvers.
It doesn't comply with speed limit laws. 🤔 Also if it did and in L4/L5 (which we likely will never have but.....) what will the car do if the navigation sends it there? And suppose it passes and the nav reroutes it around the block and back through the same route (Chuck Cook got stuck in this before). Is your robotaxi customer stuck in an infinite loop? And most important are you able to charge them by the mile or time?🤣🤣
 
First long (330 miles) drive with latest FSDb. It was great! It was terrible!

It is a great TACC + lane keep system, and it can change lanes with the press of a button on the yoke. Big roads, little roads, windy roads, hilly roads, it would slow down to a sensible speed on curves if needed, and stayed in its lane. Great; I was gruntled.
Looks like you just used AP on the highways, not FSD.
 
  • Funny
Reactions: jebinc
Looks like you just used AP on the highways, not FSD.
I don't think so, at least not for all of the trip. A combination of NoA, FSD, and a bit of AP (not sure why). Lots of good FSD stuff was happening, like obeying traffic lights, letting me go more than 5 mph over the speed limit, and random lane changes (I have automatic lane changes turned off in NoA). It definitely tried to make some turns, it just failed. In FSD's defense there were a few it might have made successfully but because of the amount and speed of traffic I decided to take over. I was so hoping it would make that one left turn, and it started out well, i.e. got in left turn lane, put on turn signal, started to turn. It maybe would have made it if I waited a bit longer, but it was also possible a speeding truck would have appeared.
 
  • Informative
Reactions: GWord and EVNow
It takes time to figure out how FSD turns and how jittery it can be. There is a learning curve to monitoring and using FSDb. You can find thousands of posts on this here ...
Yes, and there are only a few turns on the trip, most of which are either tricky, or I know will give the car (and humans) problems. For example, one is a very tight right turn that has a "no turn on red" sign. The light seems to always be red, so the car will try to turn when it shouldn't. Another is an intersection that has just been rebuilt (so not in the nav yet?) and it looks to me, other humans, and the car, like you should make a right turn into what is actually a left turn lane. The actual lane you want to turn right into can't be seen until you are in the intersection, and looks like a side street. The backroads of New York and Pennsylvania can be very fun.
 
It takes time to figure out how FSD turns and how jittery it can be. There is a learning curve to monitoring and using FSDb. You can find thousands of posts on this here ...
this is very true, the biggest difference is less about what FSDb can do and much more about an individual drivers definition of success.
To some it is a zero intervention drive if the car edges timidly round a corner in a series of stop/go's with the steering wheel spinning wildly - while others would count slowing down a fail.
 
  • Like
Reactions: FSDtester#1
That'd be something. It seems like a few years away.

I know; it's crazy, but I think Tesla is very close (from a computer programmer's perspective). Although yes, getting it all polished and working well may take longer than I think. It won't avoid potholes or be good at parking lots, but it'll drive and navigate well everywhere (not perfect).

Their progress has been much slower than I thought, but it's getting there.
 
I know; it's crazy, but I think Tesla is very close (from a computer programmer's perspective). Although yes, getting it all polished and working well may take longer than I think. It won't avoid potholes or be good at parking lots, but it'll drive and navigate well everywhere (not perfect).

Their progress has been much slower than I thought, but it's getting there.
Close to what… a usable L2 system?
 
There are a couple of areas / situations that I know 25.1 works well, and it's just so great to activate it, experience it driving smoothly, comfortably, handling everything.

We'll get to the point where it works well everywhere eventually.
The day when my car is not afraid of the cones :D . Few days back while on FSD the car wanted move away from the cones from left lane to the right lane. Except, there were cones at the edge of that lane too. Poor car was very scared :D
 
I think Tesla is very close (from a computer programmer's perspective)
to a...
A usable, comfortable, delightful, and reliable L2 system


How do you figure this? Can you be specific about how the problems with stopping and going are going to be solved in a short period of time, for example? Do you think their planner is close? Do you think they will pass lane changes to the driver to help solve that problem (I think that would be great)?

The main issues I can think of on most roads are:
1) Going
2) Stopping
3) Unnecessary lane changes
4) Bad planning.
5) Bad anticipation.
6) Bad communication with the driver on higher risk maneuvers
7) Inability to read basic signs and obey them (e.g., no right turn on red)
8) Inability to use the right speed appropriate for the road and for the circumstances.

I'm not too worried about complex city streets scenarios, special weather, etc., because those are very difficult and won't be happening any time soon (from anyone, probably, including Waymo & Cruise, though they're much closer of course!). But why is a fair weather, easy conditions, easy roads L2 assist close?
 
The day when my car is not afraid of the cones :D . Few days back while on FSD the car wanted move away from the cones from left lane to the right lane. Except, there were cones at the edge of that lane too. Poor car was very scared :D
Before I had a Tesla I never realized how dangerous and important orange cones are. My car evaluates all orange objects, and declares many but not all of them orange cones. The other day I noticed it was visualizing orange cones on the other side of a jersey barrier in a construction area. Also if you are using NoA in the presence of orange cones you are declared to be not paying attention
 
How do you figure this? Can you be specific about how the problems with stopping and going are going to be solved in a short period of time, for example? Do you think their planner is close? Do you think they will pass lane changes to the driver to help solve that problem (I think that would be great)?

The main issues I can think of on most roads are:
1) Going
2) Stopping
3) Unnecessary lane changes
4) Bad planning.
5) Bad anticipation.
6) Bad communication with the driver on higher risk maneuvers
7) Inability to read basic signs and obey them (e.g., no right turn on red)
8) Inability to use the right speed appropriate for the road and for the circumstances.

I'm not too worried about complex city streets scenarios, special weather, etc., because those are very difficult and won't be happening any time soon (from anyone, probably, including Waymo & Cruise, though they're much closer of course!). But why is a fair weather, easy conditions, easy roads L2 assist close?

All of the below is just IMO.

Stopping / going has not been much of an issue on 25.1. I did mention the 2-3s delay in going at times.

The main reason I think Tesla is very close: almost *all* of 25.1 un-delightful issues stem from:
1) poor lane choice / planning
2) unreliable left turns from the outside lane
3) inability to skip lanes when it matters
4) planning 2-3s delays

The driving itself is decent.

If you look at the list of issues above, they basically all relate to lane planning and choice. And yes, maybe Tesla will take years to solve this problem. Maybe it'll be largely solved with V11.

For me, it's all contingent on Tesla solving the lane planning issues. After that, further polish will only make the driving experience more delightful.

I know this is funny, but I think the bar for generalized L2 is much much higher than L4. People driving their own cars with their personal driving styles will be more critical of a L2 system than if they were in the backseat of a L4 car.
 
to a...



How do you figure this? Can you be specific about how the problems with stopping and going are going to be solved in a short period of time, for example? Do you think their planner is close? Do you think they will pass lane changes to the driver to help solve that problem (I think that would be great)?

The main issues I can think of on most roads are:
1) Going
2) Stopping
3) Unnecessary lane changes
4) Bad planning.
5) Bad anticipation.
6) Bad communication with the driver on higher risk maneuvers
7) Inability to read basic signs and obey them (e.g., no right turn on red)
8) Inability to use the right speed appropriate for the road and for the circumstances.

I'm not too worried about complex city streets scenarios, special weather, etc., because those are very difficult and won't be happening any time soon (from anyone, probably, including Waymo & Cruise, though they're much closer of course!). But why is a fair weather, easy conditions, easy roads L2 assist close?
And lest us not forget camera and camera placement is, well… not optimal.