TRANSCRIÇÃOEnglish

Donut Lab Solid State - The Internet Did it's Thing...

29m 14s5,614 palavras324 segmentsEnglish

TRANSCRIÇÃO COMPLETA

0:00

Donut Labs test six just came out and it wasn't even a test. It fell on April Fool's Day. We kind

0:05

of didn't know what to expect. Turned out to be an interview that was a little telling. We'll

0:08

get into that, but we kind of expected this to happen. So, I spent the last week researching

0:13

everything that the internet has been saying, all the different validation sites, all the theories

0:18

and claims and discovery and put it together in this document. So, we'll start really quickly

0:23

with what was covered in the interview and then jump right into everything we found. This might

0:28

be the most comprehensive report we've ever made. I'm Ricky and this is Two Bit Da Vinci.

0:36

We scoured Reddit. We tracked all the different sources that we could find and we have stuff from

0:41

a team called TGD that put together a 190 page report that we were actually quoted in several

0:47

times which is pretty flattering. And we've also looked at all the different theories out there

0:51

and stuff. But first, the CEO interview really quickly. Here's what he actually had to say.

0:56

one, he announced a second battery, a version two battery, which is way better,

1:01

right? When we're waiting for all the data for the first one, they have a second battery. Now,

1:04

this kind of this kind of bugs me because it feels like that classic like shell game where

1:09

you're waiting for something and then there's something even better on the horizon. Two,

1:13

he confirmed on camera that the 100,000 cycles was never measured. He quotes, "Of course not.

1:19

We would have had to literally start testing it a decade ago. It was a projection from his words,

1:24

a much smaller number." So test it X times and then extrapolate out. That's all that

1:30

was done. Pretty expand. I'd expect nothing less. I mean, I wouldn't have imagined that

1:34

they've tested to 100,000 cycles. Three, he claimed columic efficiency above one. Then

1:39

immediately caveed by saying there could be other signals showing capacity issues underneath. Now,

1:44

we've covered this in previous reports and stuff, and there's there's a little bit there, but we're

1:48

going to skip that for now in this video. Four, he again dodged the energy density question. He says,

1:53

"You never ask a woman her age and you never ask a battery it's weight funny." But he did let slip

1:59

that they've already recorded the energy density video. So hopefully that's coming up in a future

2:03

test. And of course there was some funniness. There was some Finnish humor I think in this

2:08

video. There's a person off screen saying that engagement metrics require continued ambiguity. So

2:13

this is the marketing machine at play and that's fine. So five, he was asked how much of this was

2:18

true cuz again it's April Fool's Day. What if B2 is just an April Fool's joke? The ambiguity

2:23

clearly is part of their strategy. And then they went on to sell merch. Now, that might not be that

2:27

interesting on its own. But one of the funniest things, and that's an homage to Ryan from Zuroth,

2:32

our very good friend, is a tinfoil hat. They have a tinfoil hat for sale on their store. Did you

2:37

know that? It is a hearkening. He didn't mention him by name, but he said a very famous YouTuber

2:42

already was wearing a tinfoil hat in one of his episodes when talking about donut. So,

2:46

now they're selling merch. Honestly, if it wasn't for what I've been working on for the past week,

2:49

I would have skipped this video. I've kept it very scientific until now on purpose. I didn't want to

2:55

get into this, but when there's no test data to review, well, I guess we'll have to do it. Number

3:00

10, the factory. So, this is where the world's most advanced battery is going to be manufactured,

3:05

right? Donut Lab's investor letters obtained by Finnish broadcaster YLE promised gigawatt

3:10

hour production capacity and scaling from 1 to 100 production lines. Donut Lab told

3:15

Misco Electric at CES, our very good friend, that they're at 1 gawatt hour of capacity and scaling

3:20

up. Reddit and Misco Electric traced the actual production to Nordic Nanog's facility in a mantra,

3:26

Finland. Reddit user Ona Liquid Rock drove past it and posted videos. So, this is the active video

3:31

that he posted. We'll put links to his his videos as well, but this was a couple of months ago in

3:37

the heart of winter. And there it is off here in the distance. You can see right here. Now, this is

3:42

interesting because it's a former Laplandia border duty-free shop that went vacant when

3:47

Finland closed the Russian border in 2023. The municipality invested €400,000 for a 25% stake.

3:53

Now, based on Nordic Nano's financials, it's been assumed that this would be their production line.

3:59

This test series has been incredibly fun to research. And when it comes to research,

4:03

there's a new tool I think you're going to love. This is Mammoth. And instead of paying 20 bucks

4:07

a month for Gemini, chat, and claude, Mammoth is a platform that gives you access to all the best

4:11

models in one place, whether it's for text, images, videos, research, or reasoning. The

4:16

interface is clean and familiar. But check this out. When I was researching this script, I use

4:20

Perplexity. But I can easily reprompt my request and pick a new model like Claude Opus 4.6 and

4:25

compare the results side by side. For images, Nano Banana is really good, but I can easily compare to

4:31

others like stable diffusion with the same prompt. Now, Mammoth even supports video models like VO3,

4:36

Grock video, and Sora for high-quality videos for your next project. Create custom Mammoths

4:40

with unique system prompts for the things that you do most. I have a custom mammoth for research and

4:45

another for designing thumbnails. The user prompt is a part we type in chat. It's the what, but the

4:50

system prompt is the why and how. It tells the AI how to act and think, like, be my lead researcher

4:55

or you are my chief software engineer. And now you can share your custom mammoth. So, create once

4:59

and share with your friends or team. If you like command line AI tools, there's Mammoth Code that

5:04

installs just like codecs for cloud code right in your terminal. And they even have API keys if

5:08

you're building an AI powered app. So level up your AI game, bringing the power of the latest

5:13

and greatest models in one place using Mammoth AI and get started today using my links in the

5:17

description. Huge thanks to Mammoth and you. Now back to the show. Now, why does that matter? Many,

5:22

many great companies have started in garages, right, in very humble beginnings. But CL has

5:27

100,000 employees. QuantumCape spent 2.4 4 billion and still hasn't reached full mass production.

5:33

When investor letters say gigawatt hours and the reality is three employees in a converted

5:38

retail building, that gap is big enough to bring up. We've toured battery lines at QuantumCape

5:45

and at LG in Michigan. We're talking billiondoll investments in the cleanest rooms you've ever seen

5:51

with the most controlled environments. It's a very difficult thing to make batteries. Now, again,

5:55

none of this is exactly confirmed. This just comes from like financial reportings and different deals

6:01

that people have pointed out online. I'm not making any claims about validity or truth. I'm

6:06

just going through the list of things that people have found out. Okay. Number nine, the fine print.

6:11

Donut Labs website says full charge in less than 10 minutes and a CES tomo said zero to full in

6:16

as low as 5 minutes. The VTT data says 7 minutes and 57 seconds. It's all over the place. But the

6:22

point is it is difficult to nail down exactly what manufacturers mean. You've all seen, oh,

6:27

you know, full charge in less than five minutes. Do you mean zero to 100, 10 to 90, 20 to 80?

6:33

Everybody does this a little bit differently. A common standard is probably like 10% to 80%

6:37

because the last 20% you probably don't want to supercharge if you're on a road trip. Donuts tests

6:42

were still pretty impressive. They held that full 11C for quite a bit, but the average rate

6:48

over the entire cycle was 7.5 C. But here's the bigger gap. We went through all five BTT reports.

6:54

every discharge of the healthy cell. Every single one. Discharge. Was it 1 C or 0.5C? The slowest

7:01

rate possible. Test five did discharge at 5C, but that was on a damaged cell with a broken pouch as

7:07

a safety demo. And it lost 55% capacity. If you're buying a Verge motorcycle, the the first customer

7:13

who will actually have these cells, you care about acceleration. You want a motorcycle to be fast.

7:18

That discharge rate matters. And that hasn't been tested very much. So, it was a data point

7:23

I hadn't really noticed until I started digging into some of these numbers. But charging speeds,

7:27

yeah, they've done that very well, but not this charge. Number eight, the detective. Now,

7:32

this is some of the best internet sleuththing I've ever seen. User Rectator on Reddit discovered a

7:38

Spanish company called Santa Energy making claims identical to Doughut Labs and found references

7:44

to two SGS German test reports on their website. Reddit user Signz tracked down the actual PDFs on

7:51

a partner website belonging to Next Eco CToding AG sales arm. Now there's two reports in April

7:57

and December of 2024 on a 24 amp hour cell. Same capacity class is supply chain as Donut Labs VTT

8:04

cells. Now that December cell is where the mass data comes from. Remember that number for later.

8:09

Then Signz contacted SGS corporate security directly and the April report confirmed it

8:15

was true. So the April report was genuine. The December report SGS flagged as not an original

8:21

SGS document. Someone had anonymized names post issuance breaking the digital signature. Now

8:26

this matters because in the April report on the eyewitness list, Ernst Holesenbine from CToding AG

8:33

is sitting in the same room as Robert Erman from Holivolt. The strongest physical evidence linking

8:38

these companies to the same cell technology. This matters because the theory is that Holy Volt

8:43

was licensing some kind of a technology from CT codings AG and that was what Donut is now using.

8:49

But then that all kind of fizzled and they broke the joint venture and Donut Lab stepped in. Again,

8:54

this is uncorroborated and not verified. But Donut Labs has never acknowledged where their technology

9:00

comes from. Their own investor letters say acquired access not we developed. Reddit traced

9:06

the supply chain the company wouldn't disclose and authenticated the documents that prove it.

9:11

So that is interesting. This is the power of the internet. The internet can be good. It's not just

9:16

all doom and gloom. Number seven, the deletion. On February 24th, Sana Energy's director,

9:23

Javier Cool, appeared on Donut Lab's Reddit page as Hyo Sana username. For three hours,

9:29

he answered questions openly. confirm CT coding AG as the IP holder of the technology behind Donut

9:34

Lab shared cell dimensions posted a 100 cycle chart and when pressed on energy density admitted

9:40

the battery had somewhat more than 300 W hours per kilogram not the 400 that donut claimed within

9:46

hours every comment was deleted every LinkedIn post was scrubbed and Santa requested the entire

9:52

Reddit submission be removed the moderator's take was this very forthcoming cheerful guy

9:58

got told by his very secretive bosses as DT coding to remove the information. On LinkedIn,

10:03

Santa had posted the same SGS report claiming 452 watt hours per kilogram. The report measured 297,

10:10

which is a 52% inflation on the numbers. And this wasn't isolated. Nordic Nano erased their product

10:17

page, the one listing 50,000 plus cycles and 400 watt hours per kilogram. When Reddit user Olgar

10:23

Mans found an ESA presentation with CT codings logo on Nordic Nano slides, it was taken down

10:29

and user Rectator warned, "Be careful with posting materials not seen before online. Every time the

10:35

community surfaces of connection, the evidence gets cleaned up." Now, one deletion might not

10:40

mean all that much, but this is kind of a pattern. We're seeing deletions and removals across Reddit,

10:46

LinkedIn, corporate websites, ESA presentations, and web archives. and that might tell a different

10:51

story. Now, again, I didn't see this actively happen. I didn't see the information and then

10:56

refresh and see it disappear. This is just what's been reported. I'm not positive, but it's one of

11:01

the interesting stories that I've come across that has a ton of comments and up votes and activity.

11:06

Number six, the redaction. Now, this to me is the biggest concern, and this one I can verify

11:12

because I actually looked into it myself. In Donor Lab's episode 2 video at exactly 51 seconds they

11:18

show the VTT report on screen. Six differences on page 4. Page 5 6 7 8 and 9 and 10 are identical

11:25

and then one difference on page 11. So on the left here the video draft is what they showed on screen

11:30

in their video and this is the published PDF that we've all been looking at. On screen their report

11:35

said conduct independent performance tests on the energy storage devices. What was published?

11:41

conduct independent high temperature discharge performance tests on the energy storage device.

11:46

So from plural to singular. We'll talk about why that's important. Change two. Identified a solid

11:52

state battery cells. The customer provided three visually identical cells for testing. And what

11:57

was in the published document says identified as a solid state battery cell not cells. Three visually

12:04

identical cells were provided for testing and labeled DL1, DL2, and DL3. Change four. This

12:09

is removed. One cell was subjected to charge performance tests reportedly separately and

12:14

another cell was subjected to low temperature discharge tests reported separately. The same

12:20

cells were used for high temperature and low temperature discharge tests. Finally, each

12:26

cell was subjected to different tests conducted in parallel, all of which began with an initial

12:30

capacity test. This report presents the results of the high temperature discharge test performed on

12:36

cell DL2. Just to bring this all together on page 11 there's one change to keep the singular and

12:42

plural uniform the devices and cells go to device and battery cell singular we have been asking for

12:49

that cold weather test because in engineering you want to put bounds on a problem right if

12:54

you wanted to know is somebody old or young well how what is the youngest person who is the oldest

12:58

person then I can tell you if I don't know who the oldest person is 34 is that old I don't know how

13:03

old are the oldest people so that high temperature and low temperature test would have given us that

13:08

bound. We don't have the low temperature test. And now we know that they actually did it. In this

13:13

report right here, you can see it says another cell was subjected to low temperature discharge

13:18

test. It's right there. It's right there in the original report. This is not theory or conjecture.

13:23

This is in the video from their test, too. But the final report removes it. Now, when I saw that,

13:31

I was really mad and I wanted to make sure that we weren't being duped because if you can't trust the

13:35

report, then what are we doing? I've been doing what exactly, right? It does sound to me like

13:40

there was a cold weather test and they've either removed it and they're going to share it with us

13:45

later or the high temperature test destroyed the cell and the cold test couldn't be done or the

13:51

test result that they got from the cold weather test was not favorable and they're decided not to

13:56

show it. It's one of those three things. And I'm also a software engineer, so I'm doubly OCD about

14:00

this. But when you're going to document, you have to have version control. Revision A, revision B,

14:05

right? Version two. Why? Because if I sent out a preliminary report that someone is using and it

14:11

has the number 5.2, but then I made a mistake and I fixed it to 7.2 and I didn't change the revision

14:17

number, that person might operate with bad intel. They might think that a number in a report is

14:22

there, but not be correct. So that's why if you ever look at like code compliance or anything

14:27

else, you'll always see like Rev A, Rev B, Rev Z, Rev 52. This is really bad science etiquette. This

14:33

bugs me because they've changed their report for some reason and not done that. Now, I didn't have

14:39

enough time to do the same kind of everything on screen and not for every single time they've shown

14:46

something. So, anybody in the comments, if you've done the same exercise to compare what was in the

14:50

report to what was on screen and you see any other discrepancies, let us know in the comments. But

14:55

this does bug me. I have to give credit to Reddit user Siron QLED for spotting this first. When he

15:01

mentioned it, it sounded like it couldn't be true, so I had to verify it. And yeah, they did change

15:05

it. So, why this matters is just trust, right? The trust in the report. If other things have been

15:11

doctorred or they've changed things, then we've got nothing. Like if we can't confirm something,

15:17

we have nothing like a ground truth, then this is kind of a a moot point. Number five, the cutoff

15:24

VTT test one, the safety system cut off at 11C charging at 90° C. That's the standard lithium

15:31

ion cutoff. And this could have just been a lab test limitation that they didn't change,

15:37

but our friend Ryan Enis from Zeroth who has a PhD in this stuff, he mentioned that that was one of

15:43

the first things that told him this might be a lithium ion battery. So that 90° C safety limit,

15:47

was that imposed by VTT as just how they operate or was it imposed by Donut? We don't really know.

15:54

Donut says that they didn't impose that. But VTT test 2, the cell discharged at 100 plus degrees

15:59

CC and the pouch lost its vacuum. Right? The seal was breached. In test 5, the CEO explained, "These

16:05

industry standard pouch materials are simply not designed for the 100° C that our batteries can

16:11

take." Now, the community tested that claim. John Sullivan heated a 5-year-old off-the-shelf NMC

16:17

cell to 100° C for 30 minutes. The pouch survived fine. Reddit user Reddit Mutter cited IEC62133,

16:25

a major international battery safety standard which requires surviving 130 degrees Celsius

16:30

for 10 minutes for certification. Now, ready user Mixmaster INI, who works with solidstate

16:35

batteries, says standard pouching material is stable well above 100°. We use it at around

16:40

120 for days. I'm not sure about that, but why would a Donut Lab cell fail where a 5-year-old

16:46

generic cell didn't? Either way, they used worse packaging than a commodity cell. In our research

16:52

for test two, we came across some data that helped us understand the pouch laminate systems. We don't

16:57

know for sure how they did it, but what typically is done is the outer layer is a nylon and it can

17:03

withstand 220° C temperatures. Then there's an adhesive polyurethane. It's a cross-link and

17:09

it doesn't melt. It's it's rated to bond the nylon to the aluminum and that's also highly

17:13

rated. Number three is a barrier. It's aluminum foil rated for over 600° that wouldn't fail. So

17:18

the weak link in this train is number four, the adhesive number two, which is a polyolin,

17:23

and that melts at 85.4°. So that led us to think that that might have been what happened, and

17:28

that might have been the anatomy of their pouch construction that would uh have caused problems,

17:34

but clearly that isn't a problem for other cells. We've seen other people reach those temperatures

17:39

safely. So that brings into question, how good is their manufacturing process and their knowhow,

17:44

right? Their technical knowhow. Building batteries is done by like a handful of companies on Earth

DESBLOQUEAR MAIS

Registe-se gratuitamente para aceder a funcionalidades premium

VISUALIZADOR INTERATIVO

Assista ao vídeo com legendas sincronizadas, sobreposição ajustável e controlo total da reprodução.

REGISTE-SE GRATUITAMENTE PARA DESBLOQUEAR

RESUMO DE IA

Obtenha um resumo instantâneo gerado por IA do conteúdo do vídeo, pontos-chave e conclusões.

REGISTE-SE GRATUITAMENTE PARA DESBLOQUEAR

TRADUZIR

Traduza a transcrição para mais de 100 idiomas com um clique. Baixe em qualquer formato.

REGISTE-SE GRATUITAMENTE PARA DESBLOQUEAR

MAPA MENTAL

Visualize a transcrição como um mapa mental interativo. Entenda a estrutura rapidamente.

REGISTE-SE GRATUITAMENTE PARA DESBLOQUEAR

CONVERSAR COM A TRANSCRIÇÃO

Faça perguntas sobre o conteúdo do vídeo. Obtenha respostas com tecnologia de IA diretamente da transcrição.

REGISTE-SE GRATUITAMENTE PARA DESBLOQUEAR

APROVEITE MAIS DE SUAS TRANSCRIÇÕES

Inscreva-se gratuitamente e desbloqueie o visualizador interativo, resumos de IA, traduções, mapas mentais e muito mais. Não é necessário cartão de crédito.

    Donut Lab Solid S… - Transcrição Completa | YouTubeTranscript.dev