Autopilot is useless on non-highway roads.

[Handsome Jack]

Tesla is unlikely to react - it's super niche thing, I bet there will be like 30-100 people in the states who will use that and probably a few thousands across the globe. The same way Toyotas, Hondas etc just ignore OpenPilot.
Maybe if it'll become massive tesla will do something but now we (OpenPilot users) aren't even blimp on their radar.

> I would imagine the EAP package becoming an included feature and the FSD package being lowered in price.

That would be perfect, but they want to earn money and likely won't do it. However if to look at price/value OpenPilot is much better value than EAP. Btw they also work on some sort of full self drive, a bit limited but still. Experimental mode can stop on stop signs and traffic signs etc and is in active development, possibly it'll be a new norm there.

 

[hmadsen]

Tesla is unlikely to react - it's super niche thing, I bet there will be like 30-100 people in the states who will use that and probably a few thousands across the globe. The same way Toyotas, Hondas etc just ignore OpenPilot.
Maybe if it'll become massive tesla will do something but now we (OpenPilot users) aren't even blimp on their radar.

> I would imagine the EAP package becoming an included feature and the FSD package being lowered in price.

That would be perfect, but they want to earn money and likely won't do it. However if to look at price/value OpenPilot is much better value than EAP. Btw they also work on some sort of full self drive, a bit limited but still. Experimental mode can stop on stop signs and traffic signs etc and is in active development, possibly it'll be a new norm there.

At some point they will need to make atleast some functions in their EAP and FSD packages included.

Right now, where i live, the ONLY features you get if you nuy FSD instead of EAP are the stop at stop signs and traffic lights, the "FSD" is still "Coming in the future"

It would be insane here to pay for the FSD package instead of EAP.

The reason i say they will need to make atleast some of the features included is, other car manufacturers have these as standard or will have them as standard in not very many years.

Tesla will have to invent something even more futuristic to ask for payment for.

Most every feature that was once something that had to be included at some cost will become a standard feature.

I remember back when adaptive cruise control, ESP, Lane assist and more than 2 airbags were a paid "Safety" upgrade to the car and in the beginning only the most expensive cars had those. Now every car sold (Atleast here....Law mandated) has ABS, ESP, 6 Airbags etc etc.

 

[primedive]

Ah, thanks for clarification! This why probably some people saying they can override and another don't

In EU it allows to override only on roads with 2 or more lanes in the same direction mostly out-of-town (with a few exception). On 2 lanes roads no overriding.

What you said looks like a reasonable compromise.


Yeah, that's very good summary. Luckily there is a solution now, for those who don't want or physically cannot get FSD.

Is it known why the overrides aren't allowed? Is a regulator forcing Tesla and only Tesla to do this? Other manufacturers let you set whatever speed you want even with their equivalent of Autosteer active.

 

[hmadsen]

Is it known why the overrides aren't allowed? Is a regulator forcing Tesla and only Tesla to do this? Other manufacturers let you set whatever speed you want even with their equivalent of Autosteer active.

I dont think regulations plays a role here, but in general Tesla functions is a dumpsterfire in the EU, it seems to me Tesla has a big setup program with all functions down and all countries across and then they enable or disable these functions for each country. Makes good sense, but it seems they have an orangutang that picks at random to enable or disable these functions.

Auto light on with wipers - Not enabled (Other cars i have, have had this for 10+ years so not against regulations)

Sign reading on motorway (Disabled in Denmark (And other countries i think) but works fine in Germany)

But hey, i get updates to zombie slayer game so who am i to complain

 

[Handsome Jack]

Is it known why the overrides aren't allowed? Is a regulator forcing Tesla and only Tesla to do this? Other manufacturers let you set whatever speed you want even with their equivalent of Autosteer active.

It looks like it's tesla's decision, other car makers doesn't force speed limits, they read it and can set speed but you can disable that behavior and/or override it. I bet Tesla is afraid of lawsuits or something.

@hmadsen I agree with you, but I think for the general population driving assist features aren't just that important. It is for me and apparently for you. Wondering if there is a statistic about that somewhere. Perhaps Tesla in the future will do something like offering FSD as a standard or have it cheap enough like premium connectivity so you're implied to have it with Tesla, as people will point out that basic assistance is years behind competition. Hopefully


Btw I'm using OpenPilot almost for a week and still super happy with it!

Few more thing how it's superior to autosteer:
- it can steer on roads without road markings! And reasonably well, unless it's super narrow road - it holds to the right side of the road and just drives without any stress
- it's less nervous about being in the absolute center of the lane, so if a car parked taking part of the lane, OP gracefully go around it within the lane. Autosteer in that case would panic and slam on the brakes and sound all alarms

 

[BitJam]

Ah, thanks for clarification! This why probably some people saying they can override and another don't

In EU it allows to override only on roads with 2 or more lanes in the same direction mostly out-of-town (with a few exception). On 2 lanes roads no overriding.

My free trial started today with FSD 12.3.6. It forgot how to get to the grocery store which was a bummer. It's speed limit maps are still suicidally bad in some areas, like my home neighborhood. It wants to go 25+ mph where it should only be going 15 mph tops. There is a setting that lets the car decide what speed to go based on the conditions it sees, not the speed limits. I tried this. I had it set to chill/chill and it still wanted to go dangerously fast. AND the right scroll wheel did not affect the max speed!!!!!

I changed the setting so the right scroll wheel worked and I could inhibit some of its maniacal tendencies. I also tested using the scroll wheel to set the speed faster than the speed limit. It appeared the in-town limits no longer existed but I didn't do extensive tests. I could certainly adjust the speed higher in some places where I wasn't allowed to under AP.

IMO the aggressive speed limit thing is bonkers. It wanted to drive through a parking lot at 35 mph. Maybe it assumed the parking lot was an extremely wide section of road. To be fair, I saw a Door Dash human driver trying to do the same thing.

Despite the flaws, I was very impressed. It was far more usable than the early v11 I tried last summer. Twice in about 10 miles it stopped as it was approaching a line at a light so a car coming in from the side could get through. I try to always do this but not everybody does. It also slowed way down for a speed bump. And it can easily drive places where AP refuses to drive.

If they finally fix the map so the car can take me to the grocery store and if they add a feature to let me set a max speed in certain areas then this version of FSD seems pretty good.

 

[hmadsen]

My free trial started today with FSD 12.3.6. It forgot how to get to the grocery store which was a bummer. It's speed limit maps are still suicidally bad in some areas, like my home neighborhood. It wants to go 25+ mph where it should only be going 15 mph tops. There is a setting that lets the car decide what speed to go based on the conditions it sees, not the speed limits. I tried this. I had it set to chill/chill and it still wanted to go dangerously fast. AND the right scroll wheel did not affect the max speed!!!!!

I changed the setting so the right scroll wheel worked and I could inhibit some of its maniacal tendencies. I also tested using the scroll wheel to set the speed faster than the speed limit. It appeared the in-town limits no longer existed but I didn't do extensive tests. I could certainly adjust the speed higher in some places where I wasn't allowed to under AP.

IMO the aggressive speed limit thing is bonkers. It wanted to drive through a parking lot at 35 mph. Maybe it assumed the parking lot was an extremely wide section of road. To be fair, I saw a Door Dash human driver trying to do the same thing.

Despite the flaws, I was very impressed. It was far more usable than the early v11 I tried last summer. Twice in about 10 miles it stopped as it was approaching a line at a light so a car coming in from the side could get through. I try to always do this but not everybody does. It also slowed way down for a speed bump. And it can easily drive places where AP refuses to drive.

If they finally fix the map so the car can take me to the grocery store and if they add a feature to let me set a max speed in certain areas then this version of FSD seems pretty good.

I get. there needs to be some incentive to buying FSD, but from what i have read, it seems that the basic functions in AP that is included with every car is still performed better if you have FSD, compared to just basic AP.

I dont understand how it makes sense to have 2 different pieces of software for the same functions, it is fine, it does not do all the stuff FSD is supposed to do, but if i understood it correctly, there is a lot of AP functionality that is way worse in basic AP, compared to FSD.

I wonder if they some day, atleast makes it so, that the functions that AP has, will be the same quality as the same functions in FSD.

I have to be honest...No matter how good they make it, if they do not lower the price i am not ever going to get it, if they cut the NOA package down to 1/3 of what it is today i would probably buy it, but at the current price there is no functions that's worth the price they are asking.

 

[SidetrackedSue]

I dont understand how it makes sense to have 2 different pieces of software for the same functions, it is fine, it does not do all the stuff FSD is supposed to do, but if i understood it correctly, there is a lot of AP functionality that is way worse in basic AP, compared to FSD.

Currently (with V12.3.x) the two pieces of software don't actually do the same functions.

FSDS takes away TACC (to get it you have to un-enable FSD or change your user profile to one with FSD NOT-enabled.)

To get TACC, you have to give up FSDS, which offers better lane keeping and NoA off controlled access highways.

Neither package works close to flawlessly, for instance reliably slowing or speeding up the car when the speed limit changes if you happen to have missed the sign. Predicting the speed it chooses to use is complicated by it not reading speed limit signs properly but also ignoring mapping, so, in my case this summer, it is useless for our 3 hour drives 2x per month to a rental cottage as the route takes Highway 60 and FSDS insists the speed limit is 60, not 80kph and, if the speed is overridden by using the go-pedal or scroll wheel to get up to the limit, it drops again the next time the car passes a speed limit sign.

In AP (Autosteer and TACC) it will allow the speed limit to stay overridden (but not always - still trying to figure out why) and stick with the 80 but then again it doesn't slow for towns or sharp curves unless there is a car in front. Of course, AP is technically only for divided highways, which Highway 60 is not.

So one has to choose which software package one will use, put up with the limitations of that package, and remember which package one is using so that one doesn't accidentally forget the car won't stop at a stop light or follow the highway when there is a Y intersection.

 

[DrChaos]

I get. there needs to be some incentive to buying FSD, but from what i have read, it seems that the basic functions in AP that is included with every car is still performed better if you have FSD, compared to just basic AP.

I dont understand how it makes sense to have 2 different pieces of software for the same functions, it is fine, it does not do all the stuff FSD is supposed to do, but if i understood it correctly, there is a lot of AP functionality that is way worse in basic AP, compared to FSD

Unfortunately the reason is obvious: Elon Musk doesn't care about basic AP because it doesn't make him extra money. He harshly criticizes and fires employees at whim and harangues them about FSD. Any employee working on basic AP is liable to get fired as useless, even if they are doing the right thing.

Any change requires significant software testing and effort and he hates paying for labor.

 

[BitJam]

I get. there needs to be some incentive to buying FSD, but from what i have read, it seems that the basic functions in AP that is included with every car is still performed better if you have FSD, compared to just basic AP.

I don't understand how it makes sense to have 2 different pieces of software for the same functions, it is fine, it does not do all the stuff FSD is supposed to do, but if i understood it correctly, there is a lot of AP functionality that is way worse in basic AP, compared to FSD.

It makes perfect sense to me but I've worked in software development for many years. It is similar to what many companies do including those making phones and chips. The technical term is leapfrogging.

Even mountain climbers do something similar.

Tesla is focused on developing their flagship AI product, FSD, as quickly as possible. just like Samsung is putting most of their efforts into developing their flagship S series of phones. The developments in FSD eventually trickle back to the AP product just like features in Samsung's S series trickle back to their A series.

A similar pattern occurs with Tesla's main firmware branch and their FSD beta branch. Usually the FSD beta branch lagged way behind the main branch in non-FSD features. Eventually Tesla works to merge both branches to let everything catch up. IMO this is why there was such a long delay in getting v12 to people on the 2024.8 branch. Tesla needed to merge the latest FSD with the latest non-FSD firmware. It added a month or more to the release date. By NOT merging on every release they save a lot of time.

Keeping changes isolated is a major key to rapid and reliable development. Many modern software languages and development techniques have been created around this simple truth.

 

[hmadsen]

It makes perfect sense to me but I've worked in software development for many years. It is similar to what many companies do including those making phones and chips. The technical term is leapfrogging.

Even mountain climbers do something similar.

Tesla is focused on developing their flagship AI product, FSD, as quickly as possible. just like Samsung is putting most of their efforts into developing their flagship S series of phones. The developments in FSD eventually trickle back to the AP product just like features in Samsung's S series trickle back to their A series.

A similar pattern occurs with Tesla's main firmware branch and their FSD beta branch. Usually the FSD beta branch lagged way behind the main branch in non-FSD features. Eventually Tesla works to merge both branches to let everything catch up. IMO this is why there was such a long delay in getting v12 to people on the 2024.8 branch. Tesla needed to merge the latest FSD with the latest non-FSD firmware. It added a month or more to the release date. By NOT merging on every release they save a lot of time.

Keeping changes isolated is a major key to rapid and reliable development. Many modern software languages and development techniques have been created around this simple truth.

I get that, but i just dont really understand why he on one hand does as you say, focus on the FSD package because he makes money off of it, but on the other hand keeps the price so high that a very limited number of users actually buys it.

The software essentially cost the same, if they sell it to 1 or 100 users, so to maximize profit they should sell it at the price where they get most money no matter how many they sell.

Or to put it in a nother way.

If the NoA package was 1000 USD, then i would buy it, so right now they can get 1000 USD or they can get 0 USD if they keep the much higher price.

I dont see any usecase where the NoA package is actually worth the price it cost for any user. I mean, first of all it is only 2/5 of the advertised features, and secondly, auto lanechange is hardly worth THAT many money for anyone, except for a very small subset of "nerds" that have multi 1000's of dollars to blow on it.

 

[BitJam]

I get that, but i just dont really understand why he on one hand does as you say, focus on the FSD package because he makes money off of it, but on the other hand keeps the price so high that a very limited number of users actually buys it.

The software essentially cost the same, if they sell it to 1 or 100 users, so to maximize profit they should sell it at the price where they get most money no matter how many they sell.

I don't mean to be insulting but this is an extremely short-sighted view. Unlike most corporations, Tesla's goal is not to increase short-term shareholder value. Tesla's goal is to accelerate the world's transition to sustainable energy [etc].

Elon seldom takes the short-term view. He wants to revolution driving (and then labor) by getting FSD to actually f'n work. This is why they invested $10B in AI compute and are planning to invest another $10B this year. Tesla has also been investing in FSD by putting inference computers in all the cars they sell, basically for free, while competing against cars that don't have such massive computer power.

The previous high FSD prices were to restrict the pool of beta testers to a size their then current compute power could deal with. They also wanted responsible drivers in the pool by having people who had "skin in the game". The lower prices now reflect the massive increase in compute power they recently unleashed. Their plan is to solve FSD by throwing an unimaginably large amount of compute and data at the problem in a very intelligent way. The changes in price have nothing to do with quarterly profits, they are all about solving FSD ASAP.

Actually solving FSD will transform the world. For example, we will get back almost all the 22% of space currently used in cities for parking. As another example, all automakers will be required to provide FSD or an equivalent system on all their cars. The jump in personal productivity will be through the roof. The income generated from a fleet of Cybercabs will by astronomical. Elon recently said that if you don't believe they will solve FSD then you should not invest in the company.

Focusing on FSD is not at all about making money in the short term. They are losing money hand over fist due to their massive investments in compute. The focus on FSD is to actually solve it and thereby transform the world. I'm not certain Elon will succeed but I am certain this is what he's trying to do.

 

[Handsome Jack]

I've returned recently form the trip which is explained in the original message here, and can confirm that OpenPilot solves the issue completely!
I was able to drive with the assistance about 90% of time on OpenPilot vs 5-10% on Tesla, here is the video how it looks like - New video by Handsome Jack (i.e. Tesla thinks that it's 40km/hr and will apply it for 5-10 minutes more of driving where it seems 60km/hr and applies it for the rest of trip, where actual limit is 90km/hr and sign limits speed for like 5-10 seconds only)

I can 100% recommend OpenPilot as a huge improvement to regular Autopilot. I've done already more than 1000km on it and couldn't be happier!

 

[primedive]

I've returned recently form the trip which is explained in the original message here, and can confirm that OpenPilot solves the issue completely!
I was able to drive with the assistance about 90% of time on OpenPilot vs 5-10% on Tesla, here is the video how it looks like - New video by Handsome Jack (i.e. Tesla thinks that it's 40km/hr and will apply it for 5-10 minutes more of driving where it seems 60km/hr and applies it for the rest of trip, where actual limit is 90km/hr and sign limits speed for like 5-10 seconds only)

I can 100% recommend OpenPilot as a huge improvement to regular Autopilot. I've done already more than 1000km on it and couldn't be happier!

As an aside, I always preferred European aesthetics and road design. We have nice landscapes here in North America, but we don’t have the same focus on aesthetics. Those trees would be clear cut here for sure, and with wide ditches and shoulders too. Probably some car dealerships and crazy LEDs lighting up half the sky and blinding drivers also. I would just be so happy driving manually over there I wouldn’t even need any driver assist.

 

[voyager]

Authorities and regulators have a treasure trove of valuable feedback in forums like this one.
Would anyone of them regularly study what's being written here?