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

[Knightshade]

Dan stage his videos to specially cause fsd to fail.

He also cites lots of OTHER peoples FSD videos running older versions with failures, to which Tesla folks will yell "UNFAIR! that is a version from 3 months ago!" which is the behavior to which I was referring.

Also I would imagine none HD map would have worst performance vs what is shown here with HD maps. Honestly if Tesla gets mapping right fsd is pretty much solved. Mapping is pretty much everyone's issue right now cause traffic negotiations are mostly on point.

Except then you're back to having 2 sources of truth.

Do you believe vision or the map?

If the map says there's a stop sign, but vision does not-- do you stop? Because there's cases where the sign IS there, but overgrown, and vision misses it. So which is "true"?

And how do you create the maps and keep them up to day everywhere? If your answer is using the fleet--- the fleet only has vision. So how do your maps ever end up better than just using vision in the first place? Everyone else doing HD maps is using additional sensors. I could see some usefulness for future lane selection-- but that's about it.

And my biggest lane-selection issue in V12 is not being in the wrong lane (though it DOES do that) it's how it'll bounce halfway between 2 lanes indecisively for way too long-- which definitely seems a traffic negotiation issue.

 

[Singuy]

He also cites lots of OTHER peoples FSD videos running older versions with failures, to which Tesla folks will yell "UNFAIR! that is a version from 3 months ago!" which is the behavior to which I was referring.





Except then you're back to having 2 sources of truth.

Do you believe vision or the map?

If the map says there's a stop sign, but vision does not-- do you stop? Because there's cases where the sign IS there, but overgrown, and vision misses it. So which is "true"?

Yeah that's why HD map makes self driving so easy. The question is not even asked. Also if xpeng stole from Tesla then their tech is irrelevant since it's Tesla fsd. But just watching these press videos with someone from xpeng sitting in the car, their traffic negotiations is just go to forward. Again, it's kind of how people drive in china.

 

[jamaatta]

I don't think CT helps mission right now, these are trophy cars for almost all users. They have other cars and usually an EV. We need 10 million more EVs a year. Each CT is going to eat cell production for I guess 4 25k cars. The 25k cars should take ICE off the road.

About mission: CT also helps to establish a supply chain for a 48-volt system. Which reduces use of copper tens of kilograms per car. And helps with energy effiency. It would be very difficult to start that chance with a high volume model.

 

[Artful Dodger]

several prolific accounts that worry about many concerns but don't cross the line into wildly flagrant lies/criticism, but take a negative interpretation that could be seen as reasonable by a misery guts.

Yeah, they're trying to convince HODL'ers that this is a bye-ing opportunity. This is how you know that unemployment is high in St Petersburg.

 

[Knightshade]

Yeah that's why HD map makes self driving so easy.

You skipped over the question-- when HD maps and vision disagree- which does the car believe- my stop sign situation as but one example.

The question is not even asked. Also if xpeng stole from Tesla then their tech is irrelevant since it's Tesla fsd.

AFAIK all the stuff stolen from Tesla is years old now (2018-2019 era). Well before city streets driving was a thing.

So their current most advanced offerings are not running Teslas FSD, even if they stole some foundational AP/EAP/Visualization stuff initially they developed the rest.

But just watching these press videos with someone from xpeng sitting in the car, their traffic negotiations is just go to forward. Again, it's kind of how people drive in china.

I mean, if the point is to train it to drive as local humans do-which is what Tesla tells us FSD is ALSO doing in V12- seems that's exactly what they've got.

Again I'm not suggesting their system is better-- I think that's impossible to really gauge given they each only work in vastly different places from each other. But I think it'd be pretty difficult for you to continue to argue there's "nothing like" FSD given how much like it that Chinese system is in general ODD and functionality.

 

[MC3OZ]

So how do your maps ever end up better than just using vision in the first place?

Do you find it easier to drive a familiar route, or an unfamiliar route?

IMO FSD is only ever going to work smoothly with maps. and the maps act in a way similar to human memory. The main purpose is to aid route planning and giving the car more notice of impending possible stop signs,

Whenever vision doesn't match the map, vision would normally take priority and it might trigger a map update and/or training data set update.

If multiple cars agree that a stop sign is not there then it is probably a map update, if different cars see different things, it is a training data set update.

IMO the map makes FSD a bit better than it would be with no map, that is the limit of the expectation.

 

[RSpanner]

Yeah, they're trying to convince HODL'ers that this is a bye-ing opportunity. This is how you know that unemployment is high in St Petersburg.

I think Elon's days as a visionary sooth-sayer able to get the 'I want to believe' crowd to stump up are over - at least as far as EVs are concerned.Is it in the company's best interests to pay him another $56B or hire a full-time CEO who is ideologically-attuned to Tesla's market for a fraction of that amount?Shareholders really need to lean on the board to do its duty.

 

[Singuy]

You skipped over the question-- when HD maps and vision disagree- which does the car believe- my stop sign situation as but one example.





AFAIK all the stuff stolen from Tesla is years old now (2018-2019 era). Well before city streets driving was a thing.

So their current most advanced offerings are not running Teslas FSD, even if they stole some foundational AP/EAP/Visualization stuff initially they developed the rest.





I mean, if the point is to train it to drive as local humans do-which is what Tesla tells us FSD is ALSO doing in V12- seems that's exactly what they've got.

Again I'm not suggesting their system is better-- I think that's impossible to really gauge given they each only work in vastly different places from each other. But I think it'd be pretty difficult for you to continue to argue there's "nothing like" FSD given how much like it that Chinese system is in general ODD and functionality.

Tesla engineers mention how the Chinese are using BEV net currently to do fsd everywhere. It's all still Tesla tech.

And as for the question you asked, it's HD maps. HD for a reason...you don't make stop sign mistakes. The car will stop even if the sign was blown away by a hurricane. Will humans stop? Maybe not. But the car will stop and be safer that it will stop because that intersection was designed for the car to stop. Vision will never overwrite HD maps.

 

[Knightshade]

Tesla engineers mention how the Chinese are using BEV net currently to do fsd everywhere. It's all still Tesla tech.

Do you have a citation for that? Only recent IP theft claims I can find from Tesla are around battery tech, not FSD.

I think Elon's days as a visionary sooth-sayer able to get the 'I want to believe' crowd to stump up are over - at least as far as EVs are concerned.Is it in the company's best interests to pay him another $56B or hire a full-time CEO who is ideologically-attuned to Tesla's market for a fraction of that amount?Shareholders really need to lean on the board to do its duty.

I think that depends on if you want Tesla to be just a car company (or maybe just cars and stationary storage)--- which if so then sure a stable COO-as-CEO type would be an idea replacement, to start ramping new models, get more megapack factories under construction, etc.

But the current PE ratio is probably still too high for that company so consider what that means going forward.

And I don't think that's the guy that gets you robotaxis and humanoid robots. I'm not sure who you replace Elon with that gets you things like that- and you're gonna need at least one thing like that, if not more than one, to get a market cap anywhere near what it was at ATH again.

 

[AndrewZ]

Dan stage his videos to specially cause fsd to fail. Also I would imagine none HD map would have worst performance vs what is shown here with HD maps. Honestly if Tesla gets mapping right fsd is pretty much solved. Mapping is pretty much everyone's issue right now cause traffic negotiations are mostly on point.

Will Tesla take liability if it’s possible to set up any scenario where FSD will consistently fail?

Some situations can happen very rarely but when you’re running a fleet of millions of vehicles racking up billions of miles, the system will increasingly encounter rare situations potentially resulting in bad things could lead to disastrous liability.

These tests might be relevant because liability for a generalized Robotaxi might not be taken until there is basically no fallibility, it’s like running the system through a series of worst-case scenarios. If you’re in the business of managing risk, that’s a prudent and necessary thing to do.

 

[WattsHappening]

Oh no, @Knightshade found someone else to argue about FSD with. I guess we have another 4 pages of posts to look forward to in the morning.

 

[Zedbox]

Tesla should take part of the billions they get free from the ICE companies via credits and hire a top director and make world class adverts and just post them on X. Free advertising

Bots don’t buy cars!

 

[Ellec]

Oh no, @Knightshade found someone else to argue about FSD with. I guess we have another 4 pages of posts to look forward to in the morning.

Is there a dedicated thread they could take his discussion to? It's annoying having to scroll past all their back and forths.

 

[Ellec]

https://twitter.com/x/status/1782623069825368377

Katy has over 100 million followers on X. Probably just added another year of reservations right there. Thanks Katy Perry

 

[Quesder]

China 4/15- 4/21 weeks EV insurance numbers, which is usually taken as sales/delivery number. Tesla was 6000 the previous week.
No wonder the price cut is required...
Hopefully, this is already priced in today (Mon 4/22).
BYD 57,300
问界 6000 WenJie (Huawei)
Tesla 5200
理想 4500 (Li auto)
蔚来3000 (Nio)
小鹏2500 (Xpev)
小米 1700 (Xiaomi)

On the positive side, some Tesla stores are claiming that they are selling much more cars now since the price cut....

https://twitter.com/x/status/1782653402591400399

Save

When I post the 4/1-4/7 number 1907, some people here challenged me saying a week number is nothing, and evening saying I am just spreading FUD. Now we have 3 weeks number, the trend is very obvious. However, it is too late, the SP has dropped 17%

 

[SO16]

Not good.

https://twitter.com/x/status/1782653550100693083

I’m now thinking TSLA could drop to -75% YTD. I sure hope I’m wrong. I’m tired of these “buying opportunities”.

https://twitter.com/x/status/1782634001271816516

https://twitter.com/x/status/1782463196005040278

 

[Quesder]

Because Elon expected this 2 or 3 earning call earlier. He said demand would diminish and they would still lower prices to lower or even negative margin to pump cars out. He is doing exactly as planned and what he expected is happening.

So, does he have a solution to the demand problem? Can he illustrate that to the investors?
Or before the ER call tomorrow, anyone here can guess or provide some solutions?

 

[Carl Raymond]

Something to watch while waiting for earnings. Discusses how Tesla 4680 doesn’t need to produce cheaper cells than China to still be a valid strategy. It derisks supply from China monopoly. Listening closely at the end you can almost hear Musk saying this.

Seems to solve the mystery of why Tesla built lithium refinery when all forecasts pointed to lithium mining being the future limiter.

 

[Ellec]

China 4/15- 4/21 weeks EV insurance numbers, which is usually taken as sales/delivery number. Tesla was 6000 the previous week.
No wonder the price cut is required...
Hopefully, this is already priced in today (Mon 4/22).


When I post the 4/1-4/7 number 1907, some people here challenged me saying a week number is nothing, and evening saying I am just spreading FUD. Now we have 3 weeks number, the trend is very obvious. However, it is too late, the SP has dropped 17%

The numbers you post are not FUD assuming it's accurate data. However for some investors a few weeks/months/quarters numbers going down due to economic changes in nations/regions is expected as is supply/demand price discovery by Tesla. It changes nothing about the company, there have been bad quarters, this is a bad quarter, there will be bad quarters in the future. It changes nothing about the long term potential.

If I told you Microsoft had a few bad quarter is the 2000s when Windows sold less, would you care? They are now worth over $3T. This is what these post are like for us long term investors.

Also some posts like yours come across as "Oh no more bad news im getting sick of this Elon bad, Tesla doomed, im selling", that's fine go and sell not sure why you need to announce it here, most of use are scooping up more at these prices. Selfishly I will be happy if the share price goes down further even below $100 for the next year or so as I can DCA like a king!

 

[EVMeister]

My mother just asked if she should sell half her Tesla shares (intention being to buy them back at a lower price).

The bottom is in - confirmed.