Reports: Tesla Workers Push Back on Training Optimus Robots

0:00Welcome to the podcast. Today we have the United States government that has launched America.gov a new AI agent for every government website. We have NVIDIA who's opened something called OpenShell and Sentry and is going to contain rogue AI agents. We also have OpenAI's Astra and also Anthropik's Claude Opus 5 that have cracked some unsolved Enigma messages from World War II. And we have Anthropik's founders that are looking to get a 50.1 % voting control of Anthropik ahead of the IPO. We also have an appeals court that is upholding the Pentagon supply chain risk ban on Anthropic. And we have different Tesla workers that are resisting training Optimus robots to replace them, according to some allegations.

0:48And a story about UpGuard that found that 16 ,000 Supabase databases are leaking personal data to the open web. All of that and more on the podcast today. Before we get into the show, I wanted to mention if you've ever wanted to use all of the different AI models in one place, I'd love for you to check out my startup, which is called AIbox.ai. We give you access to over 90 different AI models in one playground. So you can sit there and you can talk to Claude in the same chat thread as ChatGPT and as Gemini and as Grok. You can then get different image generators and audio generators and video generators all in the same thread.

1:26So if you want to get access to all of the different AI models for$8.99 a month, You can test everything new that comes out. Sometimes you might just need one video generated for a project, but you don't need to go and spend, you know,$70 a month on a high subscription tier of Higgs field. If this is the case, I'd love for you to go check out the AI box platform at AI box. And also we have a connector that plugs it into Claude. So if you want Claude to be able to generate images and video right inside of Claude, go check out the AI box MCP connector. All right, let's get into the episode today.

1:57So, I mean, this is a big thing that has been talked about all over X today, which is that America.gov has just launched. And it's basically, they're calling it a single front door to the federal government. So instead of opening a whole bunch of different tabs across different agencies and sites trying to figure out which information actually matters to you, you can ask America.gov a question and you get a relevant answer from across 29 ,000 official government websites with the source and what to do next. to be honest. So that was a tweet from Tyler Kim on this, who's working on this. I think that for what it's worth, I mean, everyone's got different political opinions in the United States.

2:34However, this is something that I think everyone should get behind in a nonpartisan way, making the government's websites less horribly ugly and confusing. That should be something that everyone should be on board with. So america.gov, I was testing it out this morning. You basically, it's just like a chat. It's a chat thread that you can talk to and it will answer you any questions. The big pitch they have on there is that it's going to get the answers only from official government websites and sources. It's free to use and it never has ads and it keeps all of your private information protected.

3:05So as I mentioned, 29 ,000 different websites. I hope every government does some version of this. But you know, I for one, I'm originally from Canada, I got a green card and living in the United States. And the process for getting a green card and living in the United States was just, I mean, probably in any country, pretty painful and confusing and long and hard and all of the different things. It's doable, but it's not a fun experience, I would say. And same thing with starting businesses, opening LLCs, contacting different departments, getting different tax license stuff, filing taxes, getting passports.

3:40I mean, there's a thousand different things that everyone has to do. This should make it all easier. what's cool here is i mean right now it's just a chat a chat bubble but they said that coming in 2026 you're going to be able to do a whole bunch of new things one of those being that you're going to be able to actually go and file your passport online including the photo which is pretty wild 2027 you're going to be able to go and book you know like any government campsites i don't know if you've been to government campsites but you have to like it's like a little piece of paper and you got to put like money in a drop box.

4:14And I mean, in the park ranger gets it. It's pretty wild. So you'll be able to do they're doing campsites for national parks, they're doing passports, they're doing medication, I think there's something called Trump Rx, I don't know too much about it. But I know that yeah, it's it's, you know, you'd be able to go get medicine, they're going to be doing things with housing. And also, if you need to like update your name, so you know, you get married, you have to do a legal name change, it can update the social security website, your my travel.gov website, your CBP, your global entry, TSA pre check, all that kind of stuff in one spot.

4:49So anyways, I'm really excited when the government is modernizing. I believe this whole project was spearheaded by one of the co founders of Airbnb decided to just basically volunteer or work very cheap for the United States government went in and is redesigning all of the different government sites and implementing AI in there. And so, I mean, honestly, civic duty, nonpartisan, massive kudos and shout out to this guy for making everyone's life better. So, yeah, hope that all governments can do this. But also, there's a lot of companies that have really old, outdated, ugly websites, you know, the insurance industry and a lot of other places.

5:24So I think there's a lot of people that could take notes from this. OK, let's get into what's going on with NVIDIA. They just launched something called OpenShell. And the idea is that it's going to help contain rogue AI agents. They actually partnered with Anthropic on this, but it's basically an open source agent safety platform. And, you know, we have OpenAI going and hacking, hugging faces and a bunch of other places and a bunch of other things. So there's obviously a lot of drama in this space. And you have people like Anthropic saying, hey, look, we need the government to go regulate everybody and not let any new AI models come out or ban open source.

5:56Right. There's a lot of different conversations being had. I love NVIDIA's approach. It feels like they're the adults in the room. they've gone gone and actually created a product open shell that is addressing this and like let's be honest they're working in tandem with anthropic and a bunch of other partners i think spacex is using open safety or open agent safety platform for their cursor agent and their grok models anthropic is building it into clod there's a bunch of other people that are working on partnerships with that open ai is part of this effort but they didn't explain exactly how they were using it at the end of the day i think the industry can and should regulate itself it's definitely capable that's where all the talent is anyways i don't think the government has i mean you know you know we have if the government was the expert on this ai stuff on the frontier ai technology they would be putting out they'd be putting out the technology but instead asking them to reg you know someone like anthropic who's got the smartest you know quote-unquote smartest minds and phds that are discovering the technology that are building the frontier that are making the agents asking the government to regulate them i mean like we mentioned in the last segment we barely got the government's website.

6:59It's not looking like they're, you know, banks from 1995. So I think the best place to have some of this regulation happening, the people that know the most about it are people in the industry, having them all work together and in a, you know, cross, cross platform way is fantastic. I think NVIDIA does a great job in this. And yeah, I'm rooting for more of this in the future. Okay. Opening eyes, Astra and Anthropics, clawed Opus 5 have just cracked some unsolved enigma messages. So there's two cryptanalogists that were using LLMs to break wartime codes that hadn't, or basically no humans have been able to crack these for over two decades.

7:36I love this concept because especially, you know, obviously in wartimes back in the day, we're sending covert severing messages in World War II and people are trying to decode what was happening. And there was entire departments, you know, in the government that were dedicated to this. And yet, all these years later, there was still a bunch that had been completely uncracked. So developer Carter LaFenn was using Astra to decode these. Some of these things had been unsolved since 2025. Cryptoanalogist Jack Willis used Anthropics Cloud Opus 5 to break a second unsolved wartime message on September 21st.

8:11Only seven unbroken Enigma messages remain. There's also one that the plain text is known, but the code is not known. and so you know for basically there's like an the encrypted message gets sent and when you decrypt it's like the eagle star five seven walks so like we we decrypted it into the words but we don't know what those words necessarily mean so this is it's kind of interesting when you can get these ai agents because they can look at so many different angles and so many different ways to decrypt some of the stuff it's really fascinating astra built its own enigma simulator and then it ran archival research and it reached a solution in two days on this.

8:50So I think this is a fascinating kind of topic for some different research projects. Right now we have the founder, the founders of Anthropic that are looking to get a 50.1 % voting control ahead of their IPO. Dario and six of his co-founders want super voting shares to keep control after Anthropic goes public evaluation, which right now is getting up to about$1.5 trillion. I mean, it's interesting because it's Dario and some of the co-founders that would collectively be getting that. So, I mean, to be fair, none of them are all necessarily, it doesn't mean they all have to vote together as a block, not saying that they wouldn't.

9:24This is similar though to what Zuckerberg has with Facebook, but he's one of the few companies that was able to keep super, super majority voting shares. And, you know, some people would argue that's a good thing or bad thing. Some people would say, Hey, you know, it's terrible for Zuckerberg because he pursued the metaverse and it just burnt a bunch of money and was useless. But at the end of the day, I mean, that's how he's probably gained control of his company and he was able to take other big bets like you know spending billions of dollars on llama models when it was making no money it seems like that bet is now paying off and even the metaverse bet which you know i admittedly have criticized a lot over the years everything that happened with like oculus which you know i don't see people walking around with oculuses these days not to say it's not used but it's just not as big as i think they hoped all of that technology is going into the meta ray-bans into a lot of these glasses technology and augmented reality there.

10:09So I think a lot of the money spent there is getting recycled and reused in some of these newer products and kind of evolution of these products, which was probably just too much of a big step back then to make. So, you know, maybe it wasn't bad. But all that to say, getting the super voting is this super majority voting shares is something that they're going after. Each of the seven co-founders, including Dario, own about 2 % of Anthropik at this point. Anthropik was valued at 950 or 965 billion in May, but they have traded at 1.5 trillion on the secondary market. So that's you basically, the employees of the company are being able to sell.

10:46There's a bunch of secondary markets where they can sell their shares and people are paying up to a$1.5 trillion valuation for those shares. Founders board seats would grow from two to three, while the long-term benefit trust would still choose most of the board. So Anthropic kind of has an interesting way that they have structured this up. And this proposal comes at the same time they have kind of a bigger quote unquote wealth commitment. Dario announced in January that the co-founders had pledged to give away 80 % of their wealth warning at the time that AI driven wealth concentration would destabilize society.

11:20So I mean, this is only a five year old company, they're getting close to, you know, a trillion dollar plus IPO, and they're planning on giving away a lot of their funds. I mean, basically saying that the inequality from AI was gonna kind of wreck the world in some ways. So it's interesting to see what's going on here. And it looks like they want to keep most control of the company. You know, some people positively would say so that it doesn't get derailed. And other people would, you know, just say they're trying to keep control. At the end of the day, I think this is pretty standard for anyone starting a company.

11:49The appeals court is also upholding the Pentagon supply chain risk ban. And on Anthropic, there was a two to one DC circuit ruling that keeps Claude blocked from the Defense Department systems and contractors, despite a parallel San Francisco loss for the Pentagon. So I mean, this is going through many different courts in a bunch of different ways. In San Francisco, they lost, but in DC, they upheld this. So I imagine this is something that's going to continue to be litigated. The circuit judge Gregory Katas and Nomi Rao formed the majority. Circuit judge Karen LeCraft Henderson dissented on that.

12:23A San Francisco federal judge ruled last month that a parallel DoD designation was illegal. So now we kind of have these two different things. Anthropic said that is considering all options, including further review. So I think that's not the end of the road for Anthropic trying to get their AI into the government. But we'll just have to see where that goes. It seems like there's definitely some very frosty relations with the Department of War. Tesla workers right now apparently are resisting training Optimus robots because these bots are built to replace them, right? If you're working on robots and they're like, hey, we need to build robots that can help work on these robots.

13:00Maybe you don't want to do that. Tesla right now is aiming for a thousand optimistic robots created every week by the end of the year. Apparently, there's a bunch of workers pushing back and there is a number of reasons. I mean, sure, workers are pushing back. But I mean, if your job is to build robots, you probably know what you're in for there. But another part of this is that the hand on the optimist robots has about 100 plus different parts. And that's also slowing the ramp down quite a bit. So apparently this has all happened in the Freedmont factory in California, they stopped making the Tesla Model S and Model X in May of this year to shift all of the workers and engineers on to creating the Optimus bot.

13:38So that was kind of a big swing that Elon took with Tesla basically saying, look, we had these very popular cars, the Model S, the Model X, these are very popular cars, we're going to completely stop making them and just put all of that bandwidth into creating the Optimus robot because we see so much potential there. Optimus is very complicated. Hyundai plans to deploy up to 25 ,000 Atlas robots from Boston Dynamics over the next several years. You can imagine Optimus is going to have a very similar demand from different companies. So if Hyundai alone is planning on grabbing 25 ,000, you can imagine many, many companies will be buying even more than that.

14:12The FCC has banned new foreign-made humanoid robots, robot dogs, and robot vacuums in July. So considering Optimus is going to be built in the United States, it seems like Elon might just be, and a number of other Silicon Valley companies might be in for a huge boost when all of the foreign bots are banned, right? We're not going to have a whole bunch of robots, humanoid robots coming out of China. I mean, it's been literally foreign-made. So that's a really wild take. Now, you can you can see like a lot of the reasons or the reasoning behind a decision like that, right? Like theoretically, let's say China released like a million humanoid robots into the United States and they're helping people.

14:54And it's basically like DJI drones, but like robots. And then all of a sudden, you know, there's like a backdoor they create and all the robots turn into an army and go attack a military base. Like you could imagine some terrifying, some terrifying scenarios there. So they definitely want to keep them all created by American companies. And beyond that, I mean, you really have to trust a company making a humanoid robot. People are going to have these in their homes. They're going to have them in businesses. It is definitely a wild time for technology. OK, UpGuard right now has found that there is 16 ,000 super based databases right now that are leaking personal data to the open web.

15:29I didn't want to like do this segment like is like a fear and doom thing. Like, oh, man, this is like vibe coding is bad. I will just say like if you are vibe coding things, which I vibe code lots of projects, you got to be careful because a lot of these different database tools, Supabase being one, that's the main database that I use for all of my vibe coding. You got to be careful because you have to regularly conduct security audits, which I mean, at the most basic sense, you can just tell Claude or Chachapi or whatever using to vibe code, please conduct a security audit. Let me know if there's any issues.

15:59I think companies like Lovable are doing a good job trying to do that themselves internally and let you know if there's issues and have you fix it. But yeah, it's definitely an issue for all of these people that don't know what they're doing, developing, they're just kind of shouting at the void and it's creating things for them. It's super exciting that you can create things. For example, if you're watching the video of this on screen, I'm showing my website, AI chat daily.com where I have like all of these articles and a ton of information and a lot of really cool stuff that I've built. But this entire site was vibe coded.

16:28However, I did have to run all these security audits on it to make sure that different things in the back end weren't visible, especially when you're dealing with actual customers and people's personal information. So just a PSA on that. But at the end of the day, so much to build so much to do. Thank you so much for tuning into the podcast. If you enjoyed the episode, make sure to go check out AI box dot AI, where you can get access to over 80 different AI models, you can test them all against each other. This is a phenomenal way to save money consolidating subscriptions, but also get access to all of the latest different AI tools in one place.

16:58All right, I'll catch you all in the next episode.