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

[Maarten]

If you have read my story on the previous page and think it is just about mathematics, please read again.
Because it is just as much about, in random order:
- Having a clear view of the future;
- Not being afraid to fail;
- Being at the forefront of something new and beneficial;
- Making mistakes, trying again and again and in the end getting it done;
- Not being afraid to go against real experts who will judge you;
- Not giving up when it gets really difficult;
- Not being afraid of not knowing and asking other people for help;
- Learning a lot of enriching stuff in the process;
- Not stopping where others with more means have failed;
- Working very, very hard to get something done;
- In the end reaping the benefits of all the above.

The crazy thing is: I only realized all of the above later, after pondering over it.
Maybe this, seeing it all happening at Tesla, is why I have so much faith in the Tesla people and their CEO who stimulates all of it.

 

[Thumper]

180 mile round trip yesterday on 12.3.3. Handled a roundabout perfectly. Cruising in the far left lane with traffic, our exit 40 miles up the road, suddenly for no reason at all, ego signals and moves 2 lanes to the right. It was done safely but was basically phantom lane change.

 

[JimS]

Only those who bought FSD before ~March 2019 were every promised* any autonomy (ie higher than L2) at all.

*promised as in the actual purchase page/description of what you were buying promised you actual autonomy of L4 (or arguably L5)--- not referring to any aspirational things Elon might have said at various times


Around April 2019 they began selling the "new" version of FSD with very different wording, that never promised anything about L2 with your purchase.


That said- I believe if Tesla is able to deliver >L2 to existing/legacy fleet they will do so to both groups of buyers.

But if they can not then only the first (much smaller-- and smaller still since many have since traded their "earlier version" FSD in by now) are "owed" something legally--- we just discussed that aspect in the options thread, the guesstimate at the # in this group is sub 50k at this point (based on known # of cars sold in that period, the take rate estimates from the tracker, and there being at least some % that are no longer in service or that traded their cars in)

So if I'm one of the 50k promised full autonomy, at what point does Elon buy me out of that promise? I'm easy, I'll settle for 100 shares, (as long as I can keep my car with FSD "whatever!")

 

[JimS]

Exactly, we have one of those and kept it for this reason. (FSD on a M3 RWD, spec'd at 325 mi range).

Keeping an "old" M3 just for the FSD promise? That's ridiculous. Now if you've got one with the glass roof that turns orange in the rain...that's a keeper!

 

[XPEL FTW]

Maartens- Good post. This is exactly the mindset the bulls should have, IMO. Musk, for all his warts, has beaten the odds many times. Everyone who bailed during the dark days missed out on a huge score. (I for one have doubted Musk all the way, and been pretty much dead wrong. I'm still skeptical - of the 500B+ valuation -but not with a ton of conviction any more. How can you not marvel and what Musk has accomplished?)

Most of my 'jockey' bets (Berkshire, Constellation Software, XPEL) have worked out but it takes nerves of steel; that's Maarten's post. And of course some have failed. That's why position size is vital. Make your bet, and HODL like mad. Just make sure you're not broke if it doesn't work out. That's my completely unsolicited, probably unpopular opinion.

 

[Joe F]

FSD not recognizing or blatantly breaking traffic rules in such an unsafe manner shouldn't be excused by "well the location was near by".

Quite frankly, I'd expect that the FSD system would be continuously identifying and evaluating driving conditions constantly in the background for AEB and would always have the current driving context, and "enabling" it simply switches the control outputs on.

Now, enabling it very close to a location might force it to take an awkward route (nav/GPS often does this), but it shouldn't unsafely break the rules in doing so...

Perhaps, however, the OP didn’t state when the destination was initiated, just when the route was started. A lot of variables at play, and little info supplied to evaluate the event.

I haven’t followed the technical aspects of the maps Tesla uses for a few years now, so my memory may be missing something, but I recall there is very detailed info on road “segments” including but not limited to road speed per segment as well as other info. Without access to this particular section of road map data to analyze, it’s possible that the exact position the OP initiated FSD may have been affected, at that exact point of the road, such that FSD was confused and did what it did. That’s why I suggested it should have been activated much earlier, and then if it did what it did it should be termed “bad, really bad.”

 

[primedive]

Very predictable in my opinion as the price for a Model 3 effectively jumped 15%-20% with the loss of the Federal Tax Credit. I wish Tesla would pull out all stops to make the Model 3 qualify again,

There’s also currently no referral or loyalty credits for purchasing a 3.

 

[Discoducky]

Perhaps, however, the OP didn’t state when the destination was initiated, just when the route was started. A lot of variables at play, and little info supplied to evaluate the event.

I haven’t followed the technical aspects of the maps Tesla uses for a few years now, so my memory may be missing something, but I recall there is very detailed info on road “segments” including but not limited to road speed per segment as well as other info. Without access to this particular section of road map data to analyze, it’s possible that the exact position the OP initiated FSD may have been affected, at that exact point of the road, such that FSD was confused and did what it did. That’s why I suggested it should have been activated much earlier, and then if it did what it did it should be termed “bad, really bad.”

FYI, for this issue of going the wrong way in a roundabout. Good bug!

Bug title: FSD became 'wedged' into a bad state during map metadata ingest, Severity 0, Priority 1.

Short description: There seems to be a handoff/sync issue when map nav is engaged/change while FSD is enabled.

 

[SOULPEDL]

Keeping an "old" M3 just for the FSD promise? That's ridiculous. Now if you've got one with the glass roof that turns orange in the rain...that's a keeper!

I do, but a rock hit the top glass and Tesla replaced it with clear. I was bummed, only the rear is orange now.

 

[Thumper]

I miss the orange glass.

 

[Webeevdrivers]

12.3.4 is a regression for us. Just did a short 3 kilometre trip thru town.

1. Tried to make an illegal left turn from the right lane.

2. Put on left signal to turn left at an intersection following the nav and then at last minute moved to the right lane and went straight thru signal was still on to turn left. That would have been a ticket for sure.

12.3.3 was more trustworthy for sure.

 

[JustSaying]

You and me both. But we know how people work. There is also nothing stopping Tesla from raising prices to $400mo tomorrow either.

Perhaps at some point FSD will be available by subscription only?

 

[Matias]

https://twitter.com/x/status/1779221584877928843

 

[Webeevdrivers]

Perhaps at some point FSD will be available by subscription only?

I would think so. I can’t see anyone buying it if it’s only 99 bucks a month. At least not in its present form which is absolutely no where near a robotaxi level. It would generate thousands of dollars of fines in a single day of work as a taxi.

 

[replicant]

Big FSD regression for unprotected left turns with v12.3.4

 

[2Pearls]

In other news ...the earth is flat...birds aren't real....and we for sure did not land on the moon.

I guess this note turned out to be true...problem with the accelerator pedal. From Whole Mars Catalog:

Tesla has stopped all Cybertruck deliveries for 7 days due to an issue with the accelerator pedal

 

[JustSaying]

I would think so. I can’t see anyone buying it if it’s only 99 bucks a month. At least not in its present form which is absolutely no where near a robotaxi level. It would generate thousands of dollars of fines in a single day of work as a taxi.

Pretty sure if I knew I was selling something today for $12K that I was very sure is worth $100K(?), I would make a temporarily month to month subscription price vs current lifetime purchase very attractive. We will know on 8/8...

 

[rolfrenz]

Keeping an "old" M3 just for the FSD promise? That's ridiculous. Now if you've got one with the glass roof that turns orange in the rain...that's a keeper!

Haha - got both, and am keeping it…

 

[cliff harris]

Hope the following makes an interesting story for you all.
About the difference PhD mathematicians can make for Tesla (AI?) and as a result for TSLA.

In the past I have helped someone on a master thesis at the University of Technology in Delft.
A little engineering background (don’t get scared now):
Many engineering problems (maybe also AI solving, I would be interested to know, see the following story) result in having to numerically integrate differential equations and solve n equations with n unknowns with a computer.,
Engineering students learn how to solve that during their course in numerical analysis.
A well-known universal numerical integration method that is being taught is the Runge-Kutta method.

I helped a master student to solve an engineering problem on a personal computer, the personal computer being much to the chagrin of his professor.
‘There is a big Amdahl-computer of the university that can be used, why use such a small plaything?’, the professor said.
The answer of the master student involved: ‘Because they are the future.’
And if he could pull this off, he would have a predicting model.
Well, after having programmed the personal computer it was set to work.
A few hours later, we stopped the computer and analysed where it was.
It appeared that for it to finish one run, it would take a couple of million years to complete…
Looking on the internet, we found someone in the USA that for his PhD-thesis had tried something comparable: he had the same problems, but on a supercomputer.
Clearly Runge-Kutta was not the right method.

After several futile attempts with variations of the used method, we went to consult a professor in the mathematics department of the university.
He promised to look into it and we could come back in a few weeks.
When we came back, he told us we had a problem of stiff differential equations at hand. Ehhh, what??
A small variation of a variable in one equation resulted in a huge variation in another equation, making the set very unstable to solve.
In order to solve this type of differential equations, he explained that a so-called implicit method had to be used.

So, we went to the university library and got hold of the book the mathematics professor had told us about.
We reprogrammed the computer using an implicit method, pressed run and… after half an hour we had predicting results.
At the time the model being able to predict proved to be ground-breaking.

The experience taught me the importance of having access to PhD level people in an organization like Tesla, SpaceX, Google, etc.
They can make THE difference for us as investors.
I am very happy that Tesla has proven to be a huge magnet for the most talented people in the world.
And that, with their CEO showing the proverbial example, these most talented people work unrelentlessly to get things done.

Interesting. Just to add a few thoughts on this:
1) Lets not obsess too much about qualifications. A lot of excellent academics are useless in terms of practical application. I say this with a wife who has a phd, and having met many of her associates similar qualified. Also note that Elon is very unbothered by academic qualifications.

2) On the more general point of software:
Its VERY hard for people without a software engineering background to truly appreciate how different software can be compared to everything else. In the 'real world' making something 2x as good often requires 2x the effort. Making something 10x as good is stupidly hard, and unlikely. But in software, due to the quirks of how computers operate, its very very common to find a way to re-code something and make it 1,000 or even 50,000 times faster. Some languages are way faster, some algorithms way better, and if you get to design the CPU (ASIC) it runs on, the sky is the limit,

For a long time, FSD has got a bit better, and then a bit better, and then a bit more. This has fooled observers into thinking its a linear process and like building a wall. Its not. Its not at all surprising that eventually someone found the key to 10x or 100xing the performance of it.

TL;DR: Software progress is rarely linear. FSD12 really can be THAT much better than FSD11.

 

[Krugerrand]

I guess this note turned out to be true...problem with the accelerator pedal. From Whole Mars Catalog:

Tesla has stopped all Cybertruck deliveries for 7 days due to an issue with the accelerator pedal

Meh. That doesn’t change Q1 results, nor does it change LONG TERM INVESTING. It’s likely not even going to affect Q2. So, it’s a nothing burger. Next catastrophe you or your buddy want to warn us about?