Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[Todd Burch]

With so many shared parts between 3 and Y, and even some in the Semi (like the motors), it is likely the costs of the 3 will drop even further when Y and Semi enter production.

 

[Chunky Jr.]

I could swear Elon said April was the best first month of a quarter they have ever had. I'm not sure how that got translated into a lack of demand.

 

[MarcusMaximus]

Tesla made the Model S/X refresh announcement on 4/23, which was 1 day before they released Q1 earnings which I think most here would agree were way worse than expected (even with 200m in credit sales).

To me it seems likely that the announcement was moved up a bit before the manufacturing lines and or supporting supply chain (motor manufacturing in Giga1) were ready to go. Given the huge drop off in S/X demand in Q1 that was probably a smart move.

From what we know, it’s the opposite. The drop off in production of S/X was due to lines being slowed/stopped to swap to the refreshes.

 

[Dan Detweiler]

Forbidden topic

I'll be sure to take that into consideration. Thank you so much for your kind advice. lol

Dan

 

[neroden]

OT

These things are subjective, obviously, but whatever problems there are with Tesla’s UI, it’s far and away better than any other auto manufacturer’s(or 3rd party equipment provider’s). In fact, it was just rated as such by Consumer Reports. It’s not even close.

While I can believe that, it only says bad things about the other companies. A 1970s tape deck was a more effective audio player. It's not even close.

 

[neroden]

Yes, but it's a super small team of a dozen designers or so, based at Tesla Design in Hawthorne. Keep in mind that they're not only in charge of all the MCU UIs but also the Tesla website, the app, etc.! The team size is maybe a 1/10 or 1/20 of what it actually should be, compared to FB, AAPL, GOOGL, MSFT …

Thanks for the info. I guess they've lost all the people who are good at the job, perhaps due to Musk's mismanagement.

From what I've heard, people are overworked, tired and fed up with the fact that they're reporting to either Franz or Elon – neither have much clue about UI design. The modus operandi is a toxic, top-down "We want to ship feature X so you better make it work SOMEHOW" approach.

Thanks for the picture. It does sound like a toxic, broken working envrionment with incompetent bosses.

Having reported the music player problems to Franz PERSONALLY, I certainly believe he has zero clue about UI design.

Elon clearly has no clue, since he's ordering them to waste their time on offensive easter eggs when basic functionality isn't working and regressions are appearing with every release.

Their poster boy, Andrew Kim, left in December '18 to work for Apple. He was in charge to revamp the Model S/X interface for the M3 and Semi.

Was he the one who designed the original Model S UI, which was actually very good, before it got repeatedly downgraded and broken in "updates"?

 

[neroden]

This is total nonsense. You're suggesting not only that there is zero price elasticity to all the things we import from China,

Not "all" but "most"

but that they are also a monopoly producer with no alternative suppliers globally to what they produce.

Not true for all the things we import from China...

....but true for most of the high-value items Trump is putting tarriffs on, including solar panels, solar panel components, computer chips, etc. etc. China has become the monopoly producer of a very large number of very important goods.

You need to read up on the current state of world trade. It's not what you think it is.

 

[neroden]

It's a common myth among those who don't know anything about software development that just adding more people will solve your problems. Nothing could be further from the truth. In fact adding more people beyond a certain amount can greatly slow down development and create less good solutions. Small teams of 4-8 people work best in my experience...if they are all top notch. In fact there is an inverse relationship between number of coders and time-to-delivery/quality-of-code once you get above a certain threshold.

Another thing people don't understand is that one excellent coder is worth more than 5 OK ones. The excellent one will be able to do things that none of the other 5 coders could do...in fact things that maybe 500 OK ones couldn't do. And they will do it better and in a cleaner way so it runs fast and can be more easily maintained. The job as a manager is to find these people, treat them well and keep them. Also to challenge them else they get bored. But finding them is very very hard. I am sure tesla has some of these top notch people.

I am sure they don't, at least not in the UI department (Autopilot is separate). They used to. They don't any more. Those people left. Why? How were they mistreated?

But these excellent coders are not designers usually. They are good at "designing" code, but not end user solutions. That's a different skill set and in fact is also very hard to do. Often the trick is doing very hard things but obscuring the complexity of that behind the scenes so the end user experience is clean and as-easy-as-possible. This can be very very hard to do given complex requirements. It's also the reason most software sucks...bad design.

Tesla's lost the competent UI designers. I haven't figured out who they were -- and they definitely had some really good ones back when the original Model S UI was released -- but they clearly quit. Poached by Apple? Tired of Elon's bad attitude? I dunno.

They also have clearly lost the competent coders. The number of regression bugs proves that.

As far as the "We want to ship feature X so you better make it work SOMEHOW" approach.". That's their job. Not sure what is at issue here. Management wants some feature and its the staff's job to design and implement it and to do it well. What else are they supposed to do?

Implement what the CUSTOMERS want.

There's a disconnect between management and customers right now.

 

[Skryll]

Electric Mercedes Benz doesn't scare the bejezus out of Toyota nor Detroit.

You need a compelling $25k CUV/hatchback and a compelling $35k Pickup for that.

The funny thing is, when they have that people will say: $25k still too expensive, once they have a $15k CUV and $25k pickup... goal posts always on the move.

 

[neroden]

BUT this is not the issue with the Tesla UI team, I can assure you, knowing people first hand that worked for and with them:

They are simply overworked und understaffed for the amount and scope of their work they have to deliver.

This is not general assembly where you can add a few hours of physical labour that you can sleep off the next day. Designers and engineers are productive for what, 4-6 hours net per day? Beyond that, they produce garbage and garbage only. Garbage that has to be fixed eventually. Letting them work 60h+ a week doesn't result in more amazing code or design, it results in more subpar work you can trash the other week.

Letting people work to a point where they fall asleep during mission-critical meetings – as I was told – is not a sign of great work ethics but rather of weak ass management.

There is no reason, no ****ing reason, why they shouldn't hire another 10-30 designers. There's a difference between being "spartanic" and "idiotic" and as things are right now, the Tesla UI team setup is clearly in the "idiotic" category – hence why design superstars like Andrew Kim are leaving.

Thank you again for the direct information. With the service problems being solved and the communications problems... being actively worked on, this is quickly rising to become the top problem with Tesla management.

 

[Cattledog]

Tesla/Elon really screwed up with Model 3 costs, which appear to be $38,000 for cheapest model.

I think if they would have said $35K (in whatever year they first said it) adjusted for inflation then they would have been accurate. However that is way too nuanced a message for the 20teens, where memories are short and sound bites rule. So he said $35K and then tried to hit it. I get that the addressable market increases as base car cost decreases, but I think what's out of whack is the heightened expectation of 500K that was tied to $35K.

Now Europe sort of seems to be enamored with the Model 3, so it may be a pleasant surprise.

But yeah, 400K/year Model 3 would be awesome especially if paired with 400K Model Y by 2021/22. Then at least 400K pickup and we're off to the races. I'd like to see Tesla be less sequential - let's have Model Y and pick-up by 2021. You can only be a start-up car maker for so long. Put on the big boy pants for car making and then claim to be a start-up for FSD.

 

[MartinAustin]

Awesome episode of Autoline This Week - an entire episode devoted to interviewing Rahul Sonnad, CEO of Tesloop. Discussions about longevity, financial exploitation of the car at high miles, etc.. He basically said there is no point selling a Tesla, since it can be rented out for you and make more money than it costs to own. (can you say that for any other make?)

 

[neroden]

Which would you choose?

1) A car codebase with 100% coverage of unit tests with access in the field only via the on-board diagnostics (OBD) port.

2) A code base with some sensible pre-launch tests, but not 100% coverage though the code is instrumented, the car does wireless reporting, driver instructions may be changed, and OTA updates are readily done.

I think you're making a false dichotomy. You can have 100% coverage of unit tests and access in the field over the air. Why not?

 

[neroden]

Team and business. Not the same. (One could argue Azure's success is driven by installed customer's base lock-in effect and smart business policies, rather than quality of the offer. I know I almost chose them, agains my will for a few $M/year spend. Because of how they were offering free builds and compute time, and ease of integration into our eco-system...)

But further, it's been tumultuous couple years on software self-driving side. I've hoped that team was settled by now, right people are in charge, and they have teams that works amazingly well together.

Elon feeling need to take control and shuffle things around, whether it turns out good or bad, tells me things are not great. They may be ok, good even, but not great. And that's disappointing, as software was one area to take comfort in,

Really? I was saying for years that the software side of Tesla was a total mess. Everyone here should go over to the Model S, X, 3 forums intermittently to look for the warning signs of problems with the company...

after pressure coming from production, cell capacity, margin and demand concerns. To anyone that has none of these concerns, good for you - you can't ease my concerns no matter what you say, until I see Tesla's results that debunk them.

 

[neroden]

#1 is also more problematic in a rapidly evolving code base, which I gather Tesla has. #2 can be fine but still need to test the hell outta it one way or another before going GA.

the problem for Tesla, and frankly most software projects, is that the initial code base does not consider all that will be added over the next 20-30 years because you don't know what you don't know. So if you take the quicker easier path short term your foundational architecture quickly ends up not being flexible enough so to meet the next, newest feature add-ons you end up bolting them on in what I will call inelegant ways to meet timelines instead of re-architecting to plan for expansion over time and the code base becomes shitty.

This is why I expect the entire self-driving software architecture to be thrown out and redone at least once more before we have real self-driving.

This is the bain of almost all long term code bases and interfaces. The feature explosion ends up making the user interface unwieldy and the code base a pain to maintain. Very few product architects plan well for this starting from release one onward. I have no inside knowledge of tesla to say whether they did it right or not.

On the UI, they haven't even gotten to that problem; they have more fundamental problems, like lack of regression testing and lack of functionality specifications.

Interesting and important example of lack of functionality specifications: the original Model S UI was carefully designed so that all important functionality needed while driving (a) was high-contrast for visibility, (b) had fairly large buttons for "fat fingers", (c) stayed in the same place through UI alterations so that buttons could be hit without looking. Those are specifications and they're important.

Later UI modifiers broke this.

 

[Anstandswauwau]

Having worked for Apple doing software, as well as many other top companies (Silicon Valley and elsewhere), I can state with absolute certainty that every single team was always overworked and understaffed, often egregiously so. So what?

"So what?" As if Apple's output was the standard for high quality software

Tell me, when was the last time people universally praised, technically solid, bug-free Apple software?

Snow Leopard? iPhone OS? iOS6?

Anyhow, you can count Apple's breakthrough software efforts on two hands and those micro-teams often (always?) revolved around "prodigies" like Hertzfeld, Victor, Ratzlaff and co. – which did get their fair share of extra privileges. Apple Music, Maps, infra team … yeah, not so much. And it shows.

So, yeah, understaffed and overworked teams almost always lead to shoddy output, even at Apple.

What was your point again?

 

[Maitri982]

Implement what the CUSTOMERS want.

There's a disconnect between management and customers right now.

That would be a different issue.

Really though knowing what customers want is not a problem, you know the saying about opinions are as numerous as...Tesla users are vocal and not shy about giving their opinion. Plus thousands of Tesla employees drive Tesla's and have ideas too. But what neroden wants may be very different than what evnow wants.

The trick as a manager is to figure out what in fact makes the most sense to deliver out of the sea of options before you, and that decision process is not scientific. That's why executives who can steer the ship well make a lot of cheddar.

So there's no shortage of ideas. Also, sometimes users don't know what they want until you give it to them. Designers who can come up with those things are the best.

So are you saying you didn't request fart mode as being very important? BTW... That has survived intact all new releases... No regression issues there.

 

[neroden]

It sounds like there was a cost to go from 7k to 10k. Plus, they don't have the battery supply anyways to feed Fremont to that level. As evidenced by Panasonic's recently disclosed info.

Seems like China was the smart play considering all the problems in the USA. China rolled out the red carpet for Tesla. Readily available supply of batteries in China. Chinese loans for the factory. Rapid construction for a factory they were always going to build anyways.

And China offered Musk permanent residency.

You know, if the US government and state governments continue being hostile, he might move the whole company to China, which is of course what the Chinese government actually wants.

At least they course corrected in a way that's continues their growth.

It was a fast and smart course correction, and that's one thing Musk has been very good at in the past.

 

[Intl Professor]

Perhaps they mistakenly invoked the squirrel prediction NN.

That's already overloaded in use by the ultimate CEO.

 

[neroden]

That would be a different issue.

Really though knowing what customers want is not a problem, you know the saying about opinions are as numerous as...Tesla users are vocal and not shy about giving their opinion. Plus thousands of Tesla employees drive Tesla's and have ideas too. But what neroden wants may be very different than what evnow wants.

I'd just like the USB media player to work as well as the off-the-shelf software which I can get for free on my computer. This is, seriously, the work of a couple of weeks, and I could do it myself in that time, and there are hundreds of people complaining about it. And it's just flat out embarassingly bad.

The trick as a manager is to figure out what in fact makes the most sense to deliver out of the sea of options before you,

Basic functionality comes first. Tesla's been failing at that.

So are you saying you didn't request fart mode as being very important?

Yeah, and neither did anyone else.