Speculation - if/when Tesla merges FSD to the production branch, what will that mean for AP/EAP?

[dckiwi]

I'm sure Tesla would love to throw away the old AP stack and focus only on FSD, with functionality just unlocked depending on the subscription/purchase.

It occurs to me however that this is not as straightforward as it might sound. Imagine they push a new production build in a month or so that includes FSD Beta (can be turned on if paid for), *and* every existing Tesla driver with base AP or EAP is suddenly using the new stack/visualizations.

No doubt it's a huge improvement - the visualizations alone are light years ahead of the existing AP stack. But the question is how non-FSD packages would behave.

AP: When you enable AP today you get the two blue lane lines, and the car simply does lane keeping. AFAIK it does *not* stop at stop signs or red lights. How would they recreate this experience on the FSD stack? Replace the tentacle with the same blue lane lines? Run red lights and stop signs? That seems incredibly risky. I can imagine an owner with two cars, one with FSD and one with AP. Everything looks identical about the experience, except one of them will happily run a red light with no warning.

EAP/NoAP: This one is a little more straightforward - existing owners would just have to get used to the new visualizations and driving style, but the general capabilities are the same. One difference though - when the car exits the highway, the FSD stack would need to slow down and stop and require the driver to take over, just like EAP does.

These aren't huge changes for the FSD stack, so Tesla might just implement them. They may also decide to rethink their packages as part of a full rollout. Like maybe base AP obeys traffic signals. And/or maybe they remove EAP entirely.....

 

[SidetrackedSue]

When I drive my car with the 'single stack' I have FSDb turned off and it is the same as the old AP. Behaves the same as near as I can tell (I've only driven about 150km on it so far.)

I have an FSDb profile should I be interested in using it on the fly, but I'm not, so seldom pull it up. When I do so, it is for the visualization not to actually turn on FSD to let the car drive.

I assume that for those without FSD paid for, it will look exactly as mine does. I also thought that the update was being rolled out now, not some time in the future.

 

[dckiwi]

When I drive my car with the 'single stack' I have FSDb turned off and it is the same as the old AP. Behaves the same as near as I can tell (I've only driven about 150km on it so far.)

I have an FSDb profile should I be interested in using it on the fly, but I'm not, so seldom pull it up. When I do so, it is for the visualization not to actually turn on FSD to let the car drive.

I assume that for those without FSD paid for, it will look exactly as mine does. I also thought that the update was being rolled out now, not some time in the future.

Up until now (including the latest FSD release) they have pushed completely separate software builds to FSD Beta cars. At some point, they're almost certainly going to want to stop maintaining the old AP code (and all the complexity of separate FSD Beta branches). I'm talking about that scenario, where FSD Beta just becomes a feature you can toggle on/off in the main production branch.

 

[RowdyMY]

Up until now (including the latest FSD release) they have pushed completely separate software builds to FSD Beta cars. At some point, they're almost certainly going to want to stop maintaining the old AP code (and all the complexity of separate FSD Beta branches). I'm talking about that scenario, where FSD Beta just becomes a feature you can toggle on/off in the main production branch.

Not really sure what you're asking here.

We have one with FSD(beta) and one without (AP only).

Never confusing for either of us. Ever.

If you're suggesting that someone who hasnt purchased FSD might assume that because it looks the same might try it and see, I would remind them that it's level two and they're still responsible.

It looked the same on 10.x. How is 11.x different?

 

[JHCCAZ]

Who knows exactly what they're thinking, but my guess is that basic AP will indeed be merged into the new stack. This makes the most sense from a support and UI consistency perspective. However, they might wait until they think it's ready in all markets, because they can't stop supporting and testing the old stack until they think the new one is ready worldwide.

To me that means that basic AP would gain the new visualization, there would be a unified Autosteer on highway and in-town, but of course no self-navigating Autosteer. All turns and lane changes would be manual unless forced (by disappearing lanes or construction cones etc), in which case it would probably disengage with a beep warning.

A further guess, regarding traffic lights and stop signs, is that it might improve on the older basic Autopilot by smoothly stopping without emergency braking and beeping, but proceeding from or through a green light would still require lead-car confirmation and perhaps AP re-engagement from a full stop. In other words, the same or just a notch below the current EAP Traffic Light and Stop Sign Control. The argument against this would be that it makes the standard AP "too good" as it's already a very advanced L2 in town, so would reduce the incentive of a paid EAP purchase or subscription.

If there is still to be an EAP product between basic AP and full FSD Navigate on (all streets) Autopilot, it would probably do Autopark with driver in the seat, but not the presumably impressive future driverless Smart Summon in parking lots. It would also do full traffic light and stop sign control in a through lane without requiring lead-car confirmation.

Even if all the above is right, I don't think it corresponds to the original vision of these options. The plan was that FSD really would be well along with driverless L4 by now. This speculation, about the details of Basic AP, EAP and FSD as increasingly advanced levels of L2, was not the plan - but I think they'll roll with it unless and until they get to real driverless on these cars (I know I know, just trying to respond to the question as posed).

 

[Mullermn]

I am hoping they do get merged so that we get the nicer visualisations annd also the better merging and acceleration profiles. I also think it’s likely that will happen to reduce redundancy for Tesla, and because the new stack is likely to end up safer than the old one and withholding safety improvements is quite different than holding back new features.

My concern is that the new TACC/NoA will have an active cut off coded in to it when you leave the highways/motorways where it is supposed to be used to prevent you from getting any FSD ‘for free’. That would be annoying because I currently use it on much smaller roads on occasion (under extra close supervision).

 

[EVRider-FL]

Note that Traffic Light and Stop Sign Control is not included in EAP, but is included in standard (non-Beta) FSD.

 

[JohnB007]

I'm sure the new AP will be a subset of FSD. Which subset remains to be seen, but whatever subset they need to slightly exceed the current AP capability.

IMO, doesn't make sense to go any other direction.

 

[OxBrew]

I don't see any reason to change basic AP at all, except to use the new code to do the lanekeeping. The UI can stay exactly the same.

My only concern is they start using the new code to do the TACC and lane keeping, and it's a major regression for those of us with free AP that works really well. I've not had a single PB event in the last 5k miles of highway driving, over the last 15 months. If merging everything brings back PB, I'll be really mad. Might even try to do something about it.

 

[nvx1977]

I'm sure Tesla would love to throw away the old AP stack and focus only on FSD, with functionality just unlocked depending on the subscription/purchase.

It occurs to me however that this is not as straightforward as it might sound. Imagine they push a new production build in a month or so that includes FSD Beta (can be turned on if paid for), *and* every existing Tesla driver with base AP or EAP is suddenly using the new stack/visualizations.

No doubt it's a huge improvement - the visualizations alone are light years ahead of the existing AP stack. But the question is how non-FSD packages would behave.

AP: When you enable AP today you get the two blue lane lines, and the car simply does lane keeping. AFAIK it does *not* stop at stop signs or red lights. How would they recreate this experience on the FSD stack? Replace the tentacle with the same blue lane lines? Run red lights and stop signs? That seems incredibly risky. I can imagine an owner with two cars, one with FSD and one with AP. Everything looks identical about the experience, except one of them will happily run a red light with no warning.

EAP/NoAP: This one is a little more straightforward - existing owners would just have to get used to the new visualizations and driving style, but the general capabilities are the same. One difference though - when the car exits the highway, the FSD stack would need to slow down and stop and require the driver to take over, just like EAP does.

These aren't huge changes for the FSD stack, so Tesla might just implement them. They may also decide to rethink their packages as part of a full rollout. Like maybe base AP obeys traffic signals. And/or maybe they remove EAP entirely.....

Actually I think the main problem with AP is that back in the day when ownership was low, we all knew it was only intended for highway use, and if you enabled it on a local street, you better be super alert. Over time, as AP got better people just assumed that it was intended to function on local roads with limitations like no turning or red light reaction. In reality, AP in modern times was still intended for highway only.

So if you have someone trying to enable the "new AP" that runs on FSD tech, I agree it might be shocking to see it do nothing with red lights, especially if it has the new visualization and planner line. My guess is there would have to be some modification to the visualization (e.g. no planner line) and maybe other new UI changes to make sure users know what's not supported on local streets with AP.

Or, maybe the best option is to prevent AP enabling on local streets completely. It was always considered a perk to do so anyway.

 

[MrBadger]

I can’t see it being any different to how it’s always been. Same feature sets as now, just implemented different and possibly better.

I think some people are muddying the water thinking they are going to get FSD specific features in standard AP or EAP, ie auto steer on city streets.

I can see the argument about car not stopping for lights, but regular AP is just TACC/autosteer and a few safety functions. AP/EAP doesn’t stop for them now, but I can see them doing an emergency brake in event of a collision rather than a controlled stop as FSD does now.

Likewise auto lane changes are limited now to EAP/FSD on highways, (US/Can already has this in city streets beta) but FSD will get auto lane change outside the current controlled set of routes, regulations permitted.

We already have enough precedent of feature sets, I don’t see a different implementation (single stack) changing that especially at the cost of EAP and especially FSD and personally think it’s wishful thinking to think that feature sets will be unified.

 

[Mardak]

I think non-FSD-Capability Autopilot-with-FSD-Beta will be limited to staying in the current lane and changing speed in response to the lead vehicle(s). This would mean no lane changes nor stopping for traffic control. This is still significantly better than current legacy Autopilot behavior where a curving residential road would fail to make a turn let alone the inability to activate without lines on the ground.

What gets a bit murkier is lane shifting without an explicit change such as going around obstacles, nearby vulnerable road users or parked vehicles partially in your lane, but to FSD Beta it's just trying to stay in the same lane with some offsetting. I would think a fully double-parked vehicle or closed-off lane for construction would result in the car stopping as it's explicitly a lane change: "Changing lanes away from path blockage."

I think if you were to activate this "plain" Autopilot in a turn-only lane, it would even make the turn for you to stay in the "same" lane. It might even get into the turn lane pocket as typically those are forks. But for a right-most lane that can go straight or turn right, it might ignore navigation to go straight.

As for potentially running a red light or stop sign danger, there's already an audible warning as well as physical sensation of not braking that should clue people in to not expect automatic stopping.