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

Four months to go ... tick tock

This site may earn commission on affiliate links.
Back in February, the description for FSD on the order page was changed to this:

"Coming later this year:
  • Recognize and respond to traffic lights and stop signs.
  • Automatic driving on city streets."

I know Tesla has made a lot of misleading claims about AP over the years. However to my knowledge this is the first time they put a specific date ('later this year') in the actual order page details. They are going to be under a lot of pressure to deliver on these features by Dec. 31st.

Considering they haven't released a single new AP feature since this announcement, the next four months would have to be pretty jam-packed. Could v10 actually add some significant new AP features, such as responding to traffic lights and stop signs? If not, it feels like they're running out of runway to deliver anything resembling the description above by end of year.
 
Back in February, the description for FSD on the order page was changed to this:

"Coming later this year:
  • Recognize and respond to traffic lights and stop signs.
  • Automatic driving on city streets."

I know Tesla has made a lot of misleading claims about AP over the years. However to my knowledge this is the first time they put a specific date ('later this year') in the actual order page details. They are going to be under a lot of pressure to deliver on these features by Dec. 31st.

Considering they haven't released a single new AP feature since this announcement, the next four months would have to be pretty jam-packed. Could v10 actually add some significant new AP features, such as responding to traffic lights and stop signs? If not, it feels like they're running out of runway to deliver anything resembling the description above by end of year.

Yes, I think V10 software is getting close, probably by the end of the month (but for sure by the end of the year). It feels like this will be a particularly impressive jump forward and it's going to catch a lot of the naysayers unprepared.

But it won't be Full Self Driving. I don't think FSD will even be feature complete until the first quarter of next year (and even then it won't be FSD).
 
68FC2F99-F78F-4FE9-BD67-7D2703EE73BF.jpeg
 
I know Tesla has made a lot of misleading claims about AP over the years. However to my knowledge this is the first time they put a specific date ('later this year') in the actual order page details. They are going to be under a lot of pressure to deliver on these features by Dec. 31st.
Better save a screenshot of that page.
Considering they haven't released a single new AP feature since this announcement, the next four months would have to be pretty jam-packed. Could v10 actually add some significant new AP features, such as responding to traffic lights and stop signs? If not, it feels like they're running out of runway to deliver anything resembling the description above by end of year.
I think they can probably get traffic light and sign recognition to work in some form (though probably not very reliable and dependent on map data). The "automatic city driving" is more interesting.
 
  • Like
Reactions: Silicon Desert
Considering they haven't released a single new AP feature since this announcement, the next four months would have to be pretty jam-packed.

I don't think that is correct. Tesla did release NOA without confirmation in April (?). That was an AP feature. Tesla also released the "auto steer stop light warning" feature which also could be considered an AP feature since it only works when AP is on.

Could v10 actually add some significant new AP features, such as responding to traffic lights and stop signs? If not, it feels like they're running out of runway to deliver anything resembling the description above by end of year.

I do think that we will get the traffic light and stop sign feature by the end of this year since that is just one feature that is already running in shadow mode. So there is a decent chance that it is less than 4 months away from being ready. "Automatic driving on city streets" on the other hand is pretty broad. I doubt that we will get all of it this year. We might get a small piece of it this year but more likely it will happen early next year.

There is a small chance that Tesla will dump the AP3 NN software on us before the end of the year that is "feature complete" (contains traffic light and stop sign feature and automatic city driving). But I stress small chance.

In any case, we should stop sweating about when FSD features will come out. Tesla has an internal schedule of what and when they will release. The FSD features will come when Tesla is ready to release them. In the mean time, V10 is coming. And I am sure V10 will have lots of good stuff. Let's just enjoy the ride. Good things are coming.
 
I know Tesla has made a lot of misleading claims about AP over the years. However to my knowledge this is the first time they put a specific date ('later this year') in the actual order page details. They are going to be under a lot of pressure to deliver on these features by Dec. 31st.
No there have been several "definitely happening" time lines in print over the last few years. As for the "pressure to deliver", its pretty clear that Elon doesn't bow to anyone's pressure :) and if release isn't close, then no amount of pressure can cause software to get done sooner. I've often sad that software and engineering is akin to having a baby. You won't get a baby in 3 months by putting 3 women on the job :D But you might get 3 in 9 months.
 
My best-case prediction for end of year:
  • Recognize and respond to traffic lights and stop signs. Cars will obey traffic lights and stop signs, whether on AP (lane keeping only) or NOA (Nav On Autopilot). They will stop and start without driver intervention, although they will require close supervision. Left/right turn arrows may or may not be supported initially.
  • Automatic driving on city streets. This will be similar to NOA on the highway, and will only work if NOA is available and enabled for the specific route. Vehicles will change lanes to prepare for turns and to get around slow traffic. They will also take left and right turns at intersections, although I expect many of these turns will require driver assistance.
Beyond this, I don't expect FSD to handle any complex or even relatively simple exception cases, such as: cars stopped on the road, unmarked or poorly marked roads, construction zones with detours, etc. If any of these scenarios occur, the car will either just slow down and stop or alert and hand control back to the driver.

Any NoA routes will begin after reaching a major road and conclude before vehicles reach a parking lot or parking building, requiring the user to park manually.

Even this feels like a huge stretch. I could also see the following caveats/restrictions (especially for Automatic Driving on City Streets):
  • FSD computer only. Were the above features promised to anyone who purchased FSD with a 2.0/2.5 computer? If not, 2.0/2.5 owners could be left behind for months until upgrades are completed.
  • Geofenced. 'city streets' may be worded that way for a reason. They could geofence this to major cities that are well mapped and clearly marked (although city driving is often the most complex/difficult, so this assumption may not make sense).
Of course this is all just speculation, but it's fun to speculate :)
 
My 2016 Model S with Full Self Driving as of last weeks update recognizes traffic lights when cruse control is activated. It really does not take action other than at the last possible second for hard braking - before running the red light - the instrument cluster flashes and instructs me to take control of the wheel as Auto-Pilot just deactivated. I did it a few times and there is a traffic light icon that appears on the dash. I tired the same scenario with stop signs and my car does not respond to stop signs.

Also now when exiting freeways with a red traffic light at the end of the off-ramp without a car slowing or stopped in front of me, my car will come to a full stop (sometimes).

Admittedly these are baby steps to the promise of robo-taxis, but as an automotive Engineer I find is absolutely amazing that my car benefits from over-the-air-updates and is getting better over time. All prior vehicles I have owned have essentially degraded over time (with the exception of a manual transmission shiftier that did seem to improve in feel over time).
 
Put everything into context. There are pieces that need to be in place in a certain order. For instance the car had to learn/know which lane it was in before it could do NoA. Now let’s not forget about enhanced/smart summon and what is being achieved there. To everyone it may seem like it is just going around a parking lot, but this is their sandbox where they are building the foundation for city driving. This is a big foundation that they have built which other things are being built upon.
As far as timelines go you can throw them out the window. As long as there is progress I am happy. I would rather have someone like Elon that is more worried about the safety and feature than the timeline. This isn’t something you can just rush, there are lives at stake.
 
Beyond this, I don't expect FSD to handle any complex or even relatively simple exception cases, such as: cars stopped on the road, unmarked or poorly marked roads, construction zones with detours, etc. If any of these scenarios occur, the car will either just slow down and stop or alert and hand control back to the driver.

It definitely won't do that, at least not reliably. When released, FSD will more likely hit cars parked on the road or plow through construction sites (sometimes). What you're describing would be a Level 3 system and that's not what FSD is - it's level 2 - the driver is still in charge of the dynamic driving task at all times and is expected to take over without prior warning from the car, when tricky situations arise - just like Autopilot on the freeway currently. It's exactly the same, based on the description, except for city streets.
 
Last edited:
As far as timelines go you can throw them out the window. As long as there is progress I am happy. I would rather have someone like Elon that is more worried about the safety and feature than the timeline. This isn’t something you can just rush, there are lives at stake.

I agree that this shouldn't be rushed. I didn't actually create the timeline - Tesla did, when they put 'later this year' in their FSD description. All I am doing is starting a discussion about whether Tesla's self imposed deadline for these two FSD features is realistic/likely to happen.
 
It definitely won't do that, at least not reliably. When released, FSD will more likely hit cars parked on the road or plow through construction sites (sometimes). What you're describing would be a Level 3 system and that's not what FSD is - it's level 2 - the driver is still in charge of the dynamic driving task at all times and is expected to take over without prior warning from the car, when tricky situations arise - just like Autopilot on the freeway currently. It's exactly the same, based on the description, except for city streets.

I agree that we should still treat it like a L2 system, as it can't be relied on to alert and return control to the driver. It's worth noting however that it does do this quite regularly now. In fact just today my Model 3 gave me a visual and audible alert to 'take over immediately' because it lost track of the lane markings. And NOA will often warn in advance that the driver should be 'prepared to assist' with a particularly tricky maneuver.
 
whether Tesla's self imposed deadline for these two FSD features is realistic/likely to happen

It all depends on the quality & safety desired from the feature. Personally I think they probably already have both features "basically working." But it would be fantastically dangerous to roll them out as is. "Basically working" (kind of like Autopilot) has some very significant safety problems in a city environment.

Personally, I think it's probably too dangerous to roll any of the features out until they are capable of Level 4 autonomy. Which may actually be not technically possible at this time.

In fact just today my Model 3 gave me a visual and audible alert to 'take over immediately' because it lost track of the lane markings.

Oh, I don't disagree that sometimes it tells you. But that's not really the point, is it? Arguably, that actually makes it more dangerous.
 
No there have been several "definitely happening" time lines in print over the last few years. As for the "pressure to deliver", its pretty clear that Elon doesn't bow to anyone's pressure :) and if release isn't close, then no amount of pressure can cause software to get done sooner. I've often sad that software and engineering is akin to having a baby. You won't get a baby in 3 months by putting 3 women on the job :D But you might get 3 in 9 months.

Ohmy god... the crying... whining and complaining... and then there's the kids... :) :p .

Here comes the hatemail :/