So, when is APv2 going to be noticeably better than APv1? Or is it already?

[SlicedBr3ad]

And there is a large and growing list of really small, but highly annoying bugs they could probably knock out in a week, which would make current owners ecstatic.. but that's clearly not the priority. That speaks volumes to Elon's "work etiquette and principles."

out of all your good points, i want to zero in on this one. where's the list? i actually wanted to make a post about this and to determine if anyone was able to raise their AP1/2 lurching / attempted murder / etc issues with tesla and if these got resolved.

 

[Sunshine State]

What's the forrest then? AP2? Tesla hasn't delivered on that yet, as promised in 2016.

You said "the key word is delivered."

What am I missing here?

They should work on improving existing poorly designed features and implementing others that they promised. There are students graduating every year with degrees in cutting edge software engineering. It is no excuse that Tesla is overwhelmed working on AP2 and can't spare manpower. Hire more employees if their excuse is being understaffed. Don't ignore existing sold products that were sold at high prices bolstered by a promise of continued upgrades and improvements. When AP2.5 or AP3 arrives, if AP1 history repeats Tesla will move on and AP2 will join AP1 as water under the bridge.

 

[HankLloydRight]

out of all your good points, i want to zero in on this one. where's the list? i actually wanted to make a post about this and to determine if anyone was able to raise their AP1/2 lurching / attempted murder / etc issues with tesla and if these got resolved.

You can start here, although it's a bit dated: Vote on software features/bugs to fix in 7.1 and beyond

Here's a list of USB/Media player bugs: Comprehensive USB Bug List

There are many other threads that talk about specific AP, Nav, UI, or other bugs, but you'll have to search for those.

 

[HankLloydRight]

Hire more employees if their excuse is being understaffed.

I do not think the excuse is understaffing. It's a management philosophy from the top down. If they wanted to fix all these existing problems, they could do that easily. But that's not the priority. They are in the heart of Silicon Valley -- there's no shortage of really smart people, especially ones that want to work at Tesla. But even hiring more people to work on the lower priority stuff increases the management load -- see Brook's Law again.

 

[ChrML]

I do not think the excuse is understaffing. It's a management philosophy from the top down. If they wanted to fix all these existing problems, they could do that easily. But that's not the priority. They are in the heart of Silicon Valley -- there's no shortage of really smart people, especially ones that want to work at Tesla. But even hiring more people to work on the lower priority stuff increases the management load -- see Brook's Law again.

Probably because EAP/FSD has top priority, all the talented developers gets moved onto that.

The UI developers are probably the less skilled developers, those who always do programming the hardest way or seem to always leave in a subtle bug or two for every feature they make. The manager of the UI team is probably pissed his employees can never get anything right, while all the talented people get moves into the EAP/FSD team .

 

[RDoc]

Probably because EAP/FSD has top priority, all the talented developers gets moved onto that.

The UI developers are probably the less skilled developers, those who always do programming the hardest way or seem to always leave in a subtle bug or two for every feature they make. The manager of the UI team is probably pissed his employees can never get anything right, while all the talented people get moves into the EAP/FSD team .

Having actually worked for decades in software development at many level, I would say this is complete nonsense.

People who do the kind of software needed for AP have a completely different skill set from those who would do the UI visible to users and cannot be interchanged successfully.

I'd comment further, but as they say: You'll seldom see only one fool in an argument.

 

[RDoc]

I do not think the excuse is understaffing. It's a management philosophy from the top down. If they wanted to fix all these existing problems, they could do that easily. But that's not the priority. They are in the heart of Silicon Valley -- there's no shortage of really smart people, especially ones that want to work at Tesla. But even hiring more people to work on the lower priority stuff increases the management load -- see Brook's Law again.

I totally agree and this has been the case for years at Tesla. I've never been able to figure out why Elon and Tesla are so disinterested in quality software. Clearly they are counting on it heavily for future capabilities, and as you say: they are in Silicon Valley where there are a bunch of software managers who've build organizations to do this kind of thing.

 

[HankLloydRight]

The UI developers are probably the less skilled developers, those who always do programming the hardest way or seem to always leave in a subtle bug or two for every feature they make

You're equating lack of coding skill with "programming the hardest way"... I've seen some extraordinarily skilled developers produce some of the most complex and over-engineered solutions (with *lots* of bugs) compared to junior, less experienced developers who solve problems in a much more direct and elegant way and much less code.

 

[brianman]

I do not think the excuse is understaffing. It's a management philosophy from the top down. If they wanted to fix all these existing problems, they could do that easily. But that's not the priority. They are in the heart of Silicon Valley -- there's no shortage of really smart people, especially ones that want to work at Tesla. But even hiring more people to work on the lower priority stuff increases the management load -- see Brook's Law again.

You lost me here. The rest of the paragraph is on target.

Nothing is easy in real products with real customers.

 

[ChrML]

You're equating lack of coding skill with "programming the hardest way"... I've seen some extraordinarily skilled developers produce some of the most complex and over-engineered solutions (with *lots* of bugs) compared to junior, less experienced developers who solve problems in a much more direct and elegant way and much less code.

I'm a strong believer that the best developers always find easy and flexible ways to solve the problems. Ie the best code is medium short and consise. Designed in a way that when testes properly will never cause a mishap (defensive programming with just the right amount of layers).

I've also seen seemingly skilled developers over-engineer rather simple problems, or design code that ends up in 6 layers and 10 000 lines of code, that could be solved in 3 layers and 500 lines. And even though designed to be more flexible, ends up to be too hard to maintain. I believe the reason for that is mainly cleverness/a smart mind but also inexperience. These developers usually sort themselves out with a little guidance and ends up becoming extremely good in the long run.

Luckily most high level languages are quite quick and easy to restructure and rebuild when further flexibility is needed.

 

[ChrML]

Having actually worked for decades in software development at many level, I would say this is complete nonsense.

People who do the kind of software needed for AP have a completely different skill set from those who would do the UI visible to users and cannot be interchanged successfully.

I'd comment further, but as they say: You'll seldom see only one fool in an argument.

You are probably correct.

Yet, smart minds that don't necessarily have any first hand experience with the field (AI) always seem to adapt quickly and see solutions that not necessarily everyone see. And very talented people, even though originally hired for the UI department, tend to move towards the more challenging problems.

 

[Mark Z]

I just want Tesla to emphasize keeping the car in the lane by concentrating the software routines that steer the car based on the position of the left lane line. The right lane line would be used if no left lane line exists. They should use the rear camera to double check the lane position if sunrise/sunset causes false data. While this may seem simplistic for AP2, the results would most likely be far superior to what we experience today.

 

[HankLloydRight]

ou lost me here. The rest of the paragraph is on target.

Nothing is easy in real products with real customers.

I'm basically saying they could squash the easiest 100 minor bugs in very little time, but they would have significant impact on user experience. If they have time to fart around with Rainbow Road, Mars Lander or 007 easter eggs, they could do just a little bit of bug fixing at the same time.

 

[brianman]

If they have time to fart around with Rainbow Road, Mars Lander or 007 easter eggs, they could do just a little bit of bug fixing at the same time.

You're assuming this isn't already happening.

 

[HankLloydRight]

You're assuming this isn't already happening.

Yes. Yes, I am, because we really haven't seen any attention paid to fixing bugs that affect the user experience, like the Media Player, which frankly, is one massive bug.

 

[bob_p]

Since Tesla doesn't provide any information on how they are developing their software, or their plans for the next and subsequent releases, or even acknowledge major problems with the current software (or when they'll fix those issues), all we can do is look at the released software - and make some educated guesses on what's happening.

• AP2 software is taking much, much, longer to get working than Tesla expected. This is most likely due to an underappreciation of the functionality Mobileye provided for object recognition - and replacing that capability is more difficult in real-world than what Tesla likely did a year ago in some "proof-of-concept" prototype testing.
• It appears the current Model 3 software is missing many features - no FM radio, no Bluetooth audio streaming, no Homelink, ... Since this is functionality that has existing in the Model S for over 5 years, if Tesla had invested in a solid design for the Model S/X software, adapting it to the Model 3 shouldn't have been too difficult - just adding a different user interface (in landscape mode on a smaller display). It's likely the reason why we haven't seen any media player fixes in the 8.x releases is that the software team is spending a lot more time trying to get the Model 3 software working.
• After 5 years since the first Model S was produced, the core apps are still buggy and lack obvious features (waypoints, playlists, route customization, real time traffic incident reporting, slow & limited browser, poor smartphone integration ...).
• Tesla testing process still misses obvious bugs and design flaws that appear to get through Tesla's internal and hand-picked external testers.

Taking an educated guess - these are symptoms likely indicating a combination of poor software development leadership, coupled with the lack of a solid development process & design philosophy. Even with limited software resources, Tesla should be able to have better results with their software releases.

I can empathize with the software developers. They are likely working long hours over extended timeframes, under unrealistic schedules, with changing priorities - and a lack of strong management support or solid design tools.

Among Tesla customers, there are a number of us with significant experience in managing large, complex, software projects - and doing so with consistent results - hitting aggressive schedules - and meeting quality goals. Tesla could take advantage of this relatively friendly group - and get some advice on how Tesla could improve their software results.

 

[schonelucht]

Among Tesla customers, there are a number of us with significant experience in managing large, complex, software projects - and doing so with consistent results - hitting aggressive schedules - and meeting quality goals. Tesla could take advantage of this relatively friendly group - and get some advice on how Tesla could improve their software results.

Not sure what you are suggesting here, but you probably already know, some friendly advice is not going to cut it. Tesla's software development history is a result of the internal culture that's absolutely coming from the top dog. Believe in the impossible, always cut the edge and never take time to regroup and consolidate. It's the latter issue that is biting them hard. Software doesn't inherently deconstruct. Mess piles on and it's manageable until it is not. I am not sure we are at that stage yet, but if nothing changes, we'll eventually get there. Luckily there is always hope : the lack of recent progress could mean that they are now taking the time to regroup. It'd be ironic if it's the fundamental research guy that manages to do so rather than the experienced corporate software manager but if that's what it takes, it'll be welcome.

 

[bob_p]

There is a lot of expertise that Tesla can hire (or they could tap their owner community, who'd be happy to help) to review their software processes and philosophy - and offer suggestions for how they can improve the quality of their software, while also continue to hit aggressive schedules - it can be done, but rarely happens without some planning.

This isn't "rocket science" - others have done much better - and while Tesla's doing some unique things in their software, overall others have figured out how to build higher quality software - on more reliable schedules...