The Truth Behind DeepSeek

Summary

The Truth Behind DeepSeek
The 404 Media Podcast | Released on January 29, 2025

In this episode of The 404 Media Podcast, hosts Joseph, Emmanuel Mayberg, Jason Kebler, and Sam delve into the seismic impact of DeepSeek on the AI industry. DeepSeek, an advanced AI model developed by a Chinese company of the same name, has rapidly shaken the foundations of the artificial intelligence landscape, prompting widespread discussion and concern both within and outside the tech community.

---

Understanding DeepSeek

Emmanuel Mayberg opens the discussion by explaining the origins and capabilities of DeepSeek:

"DeepSeek is an AI model that's developed by a Chinese company of the same name that itself was spun out from a Chinese hedge fund. It released last week and scored very well on the benchmarks, even outperforming OpenAI and Meta's Llama in several areas."
[01:38]

DeepSeek's impressive performance on standard AI benchmarks has fueled excitement and anticipation. As a leading model in the "frontier model space," DeepSeek's ability to surpass established giants like OpenAI and Meta with significantly lower training investments has garnered attention.

Rising Concerns and Market Reactions

The hosts discuss the growing apprehension surrounding DeepSeek's emergence, particularly its implications for the U.S. AI industry:

"The narrative in people's heads is like, China beat the US with less money and old technology. What the fuck is that?"
—Joseph [06:20]

This sentiment reflects broader fears that China's strategic advancements in AI could undermine American technological supremacy. The significant factor exacerbating these fears is DeepSeek's development using outdated chips and a training budget of approximately $5.6 million—a stark contrast to the billions invested by U.S. tech firms.

Susan Ettlinger contributes an illustrative analogy to highlight the perceived disparity:

"It's like the Americans are only getting the PS5s. The Chinese companies can only have PS3, maybe even a PS2. But they made a really sick game for PS3 and PS2. And everybody's like, damn, maybe you just don't need the latest console."
[09:13]

This comparison underscores the shock and disbelief that a Chinese company could achieve superior AI performance with seemingly inferior resources.

Impact on Nvidia and the AI Hardware Race

A significant portion of the conversation centers on Nvidia's role in the AI hardware ecosystem and how DeepSeek's success threatens its dominance:

"Ever since the AI boom took over the world, Nvidia has been the primary... the key designs and chips that are powering this AI revolution."
—Emmanuel [08:08]

DeepSeek's ability to outperform U.S. counterparts using older Nvidia chips has led to a dramatic decline in Nvidia's stock value, triggering market volatility. This development challenges the prevailing paradigm that superior hardware and massive compute clusters are essential for AI supremacy.

Industry Leaders Weigh In

Marc Andreessen, co-founder of Andreessen Horowitz (a16z), likens DeepSeek's emergence to a pivotal historical moment:

"Think it's to be seen if he's correct, but I totally understand the sentiment. He called it AI's Sputnik moment."
—Emmanuel [10:02]

Referencing the 1957 launch of Sputnik, Andreessen emphasizes that DeepSeek serves as a wake-up call for American AI firms, highlighting vulnerabilities in the current strategy of racing for larger and more advanced chip clusters.

Censorship and Ethical Implications

The discussion shifts to the ethical dimensions of AI model development, comparing DeepSeek's censorship practices with those of ChatGPT:

"The argument is that if we accept OpenAI's censorship based on their ideology, should another country create a dominant AI model with its own cultural restrictions?"
—Emmanuel [13:15]

Susan Ettlinger raises concerns about the broader implications of accepting AI models constrained by specific cultural or political agendas:

"Facebook does bad things, TikTok does bad things. We are more willing to accept the bad things that Facebook does because it's our social media company as opposed to the Chinese social media company."
[09:16]

The ability to modify DeepSeek's open-weight model allows users to bypass certain restrictions, leading to debates about information control and the potential for AI to propagate specific ideological viewpoints.

Technical Insights and Future Outlook

Emmanuel Mayberg provides a nuanced perspective on the future trajectory of the AI industry amidst these developments:

"AI is moving at the pace of scientific discovery and is developing at a rate that is very hard to predict... I am doubtful that this [market fluctuation] will hold. Nvidia is still the dominant hardware player in this space."
—Emmanuel [18:21]

He speculates that while DeepSeek's current success poses challenges, U.S. tech giants like Nvidia, Meta, and Google are likely to counteract by innovating and improving efficiency. The competitive nature of the AI race suggests that the market may stabilize as American companies adapt to maintain their leadership.

Conclusion

The episode concludes with reflections on the emotional and unpredictable nature of AI advancements juxtaposed against market reactions:

"AI is moving at the pace of scientific discovery and the markets are moving emotionally."
—Emmanuel [18:21]

As DeepSeek continues to make waves, The 404 Media Podcast underscores the necessity for the U.S. to bolster its AI research, talent acquisition, and hardware capabilities to stay competitive in this rapidly evolving field.

---

Notable Quotes:

• Joseph on market panic:
  "China beat the US with less money and old technology. What the fuck is that?"
  [06:20]

• Susan Ettlinger on censorship analogies:
  "Facebook does bad things, TikTok does bad things... we are more willing to accept the bad things that Facebook does because it's our social media company as opposed to the Chinese social media company."
  [09:16]

• Emmanuel Mayberg on AI's unpredictable pace:
  "AI is moving at the pace of scientific discovery and is developing at a rate that is very hard to predict..."
  [18:21]

This comprehensive discussion provides listeners with an in-depth understanding of DeepSeek's implications for the AI industry, highlighting the intersection of technology, market dynamics, and geopolitical tensions.

Transcript

[Susan Ettlinger]

The PC gave us computing power at home, the Internet connected us, and mobile.

[Emmanuel Mayberg]

Let us do it pretty much anywhere.

[Susan Ettlinger]

Now, generative AI lets us communicate with technology in our own language, using our own senses. But figuring it all out when you're living through it is a totally different story. Welcome to Leading the Shift, a new.

[Emmanuel Mayberg]

Podcast from Microsoft Azure.

[Susan Ettlinger]

I'm your host, Susan Ettlinger. In each episode, leaders will share what they're learning to help you navigate all this change with confidence. Please join us, listen and subscribe wherever you get your podcasts.

[Joseph]

Hello and welcome to the 404 Media podcast where we bring you unparalleled access to hidden worlds, both online and IRL. 404 Media is a journalist founded company and needs your support. To subscribe, go to 404 Media Co as well as bonus content every single week. Subscribers also get access to additional episodes where we respond to their best comments. Gain access to that content at 400, 404 Media co. I'm your host, Joseph, and with me are two of the 404 Media co founders, the first being Emmanuel Mayberg.

[Emmanuel Mayberg]

Hello.

[Joseph]

And Jason Kebler. Hey Sam. Will join us for the second half of the show. Right now there is one story which is basically across all of tech, all of AI, and Emmanuel wrote about it. The headline is Deep Seek Mania Shakes AI Industry to its Core. Emmanuel, first off, let's just keep it super basic to start. What is Deep Seek and where did it come from? Because to a lot of people it feels like this just happened overnight.

[Emmanuel Mayberg]

Yeah, DeepSeek is an AI model that's developed by a Chinese company of the same name that itself was spun out from a Chinese hedge fund. And yeah, I've heard about it several times. I follow the frontier model Space, which is a way to refer to the leading AI models and it released last week. The AI community was very excited about it. Whenever one of these models is released, there's a bunch of benchmarks that people run to see how it compares to all the other models. It scored very well on the benchmarks was beating OpenAI on some things was beating Llama, which is the open weight model from Meta. Again, this is normal. The newest thing is usually pretty good. And the hype started to build more and more as people learned more about how it was made. And now we're at the point where I'm surprised. I think like my best way to measure how impactful a particular piece of news is is when I hear about it from my mom. I almost never get a text asking what this or that is in the news. And I got a text from her yesterday which was just like, hey, do I need to worry about this Chinese AI thing?

[Joseph]

What was she potentially worried about? And I mean, I'm not trying to put your mom on blast. It's more, it's just, it's symptomatic of. This isn't just hype, there's almost like worry and concern.

[Emmanuel Mayberg]

Totally. And I think for me, as someone who's like followed it from the beginning, I was totally prepared for everything that was about to happen. Over the weekend it became very clear that this was going to have an impact that is going to escape the AI nerd community and going to have a real impact on the market. And I think that's why I heard about it from her. I don't watch cnn. She watches a lot of cnn and I'm sure that she just turned on the TV in the morning and it's one of those screens that's showing you all the stocks and everything is red and the news anchors are like, China, AI America, World War 3, question mark, you know, and she was like, oh, I'll call my son who knows computers.

[Susan Ettlinger]

Wait, but let me take a shot at expl explaining like why people are freaking out. Because it is true that this is a Chinese developed large language model. The model is called R1, the one that they released. And then the ChatGPT competitor is just called DeepSeek and you can use it on your phone. So it's the top app in the app store. So there's the sort of normal like, oh, Americans are using Chinese tech. We're worried about that. But the big thing is that this was developed using old chips and fewer chips than normally are used. And they spent some somewhere around like $5.6 million or something like that on training, which is a number that as you explain in your article, we should be skeptical of. But like the argument that the people who are freaking out are Saying is like OpenAI and Nvidia and all American tech companies have spent a ton of money training their AI models. And by a ton of money, I mean billions and billions and billions of dollars. Here is a Chinese company that comes along and spends frankly like $5.6 million if you take that at face value, which I don't know if we can, that's like peanuts for any AI startup. Like, that's essentially no money. It's performing better than OpenAI's most advanced publicly released model. I think it's the. It's O1 is, is OpenAI's most, most advanced model, I believe. And it's free. It's. It's released for free. And it's. And they did it using old chips because China cannot get the newest Nvidia chips because of export controls. So the narrative in people's heads is like, china beat the US with less money and old technology. What the fuck is that? Right, Emmanuel? Like, is that, is that the argument for, like, why people are freaking out?

[Emmanuel Mayberg]

Yeah, that's totally accurate. And the, like, the, the single headline that dominated Monday morning and what sent the markets into a frenzy is that this impacted Nvidia's value. Ever since the AI boom took over the world, Nvidia has been the primary. I mean, you could argue tsmc, which is like further up the supply chain or further down the supply chain, further up the supply chain and chip manufacturing. But Nvidia has the key designs and chips that are powering this AI revolution. And Meta, Google, Amazon, OpenAI, Microsoft, all these companies, when they're building AI, they are paying Nvidia for massive amounts of chips in order to train their AI. So Elon Musk very proudly announced last year that Xai, which is X's or, or Tesla's or Elon Musk's AI company, has a cluster of 100,000 Nvidia chips. And those clusters just keep getting bigger. And like, at the moment, all these companies are kind of eyeballing their path to, like, can we make a cluster with a million chips? Because that's going to give us a much strong, stronger AI model. And that's been the paradigm we've been in for the past three to four years. More chips, more compute, more training data equals better AI. Whoever has the best and most is going to win the AI race. And if you believe what these people say about how important AI is going to be, you're going to dominate the world.

[Joseph]

And more advanced chips, crucially, because Nvidia will release new ones every single time. And the idea is the American companies will be getting those. So it's not just quantity or expenditure, it's, as you say, the access to the latest and greatest. Allegedly greatest. Yeah.

[Emmanuel Mayberg]

And like, at the highest level, these companies are competing for, like, oh, Nvidia just announced a new design and they're competing like, okay, well, put me down for a million chips, you know, and they're finding about who's going to get them first because it's so important to have these chips in order to win the race.

[Susan Ettlinger]

I believe they've called us like a digital moat or something where it's like we, like US companies will get the, the awesome chips and China will get like old tech. And that will naturally just keep American tech companies ahead of Chinese tech companies.

[Joseph]

If I can give a ray, crude analogy to explain it, it's like the Americans are only, they get the PS5s. The Chinese companies can only have PS3, maybe even a PS2. But they made a really sick game for PS3 and PS2. And everybody's like, damn, maybe you just don't need the latest console.

[Susan Ettlinger]

I think that's it. I think that's pretty much it.

[Emmanuel Mayberg]

That's for me, totally. Imagine, suddenly a Chinese game developer releases a PS2 game that looks like a PS5 game. Everybody's like, hey, what am I doing wasting all this money, right?

[Susan Ettlinger]

And they're like, we made this in a weekend with some spare parts.

[Joseph]

I think we'll get back to the stocks in a second. But who are some of the big names reacting and how are they reacting? I'm thinking of a 16Z. Are they and what are they saying?

[Emmanuel Mayberg]

Yeah, everybody's reacting. I mean, obviously before it became world news, it was huge in the AI space. So everybody had a take. Marc Andreessen, who's one of the co founders of a 16Z, he said.

[Susan Ettlinger]

I.

[Emmanuel Mayberg]

Think it's to be seen if he's correct, but I totally understand the sentiment. He called it AI's Sputnik moment. And that is referring to the Russian satellite which got ahead of the United States in the space race. And it was kind of an oh shit moment for the United States to say we're in a more competitive space with Russia than we realized about space space exploration. And it's the Cold War, so it's the weaponization of space and all that. And that kind of accelerated the US space program. And I think that is the analogy he's making. It's a wake up call for American companies to realize that these sanctions that America, or sorry, sanctions is probably not the right word. These export restrictions on AI chips that the Biden administration has put in place for the reason of hobbling their ability to compete in this AI race, right? That's not going to cut it. Right. And the idea that we just have a massive amount of capital and we're just going to throw it at these AI problems with these bigger and bigger clusters, which has people expressed doubt about the technology and theory behind that starting last year. But this is tangible proof that that is not going to cut it. The US needs to be competitive in more areas than just capital and access to chips. We have to. Not we. The US has to have innovative research. They have to have the best talent. They have to realize that the Chinese universities and research that are happening in China are not to be dismissed. These are real competitors that can make real progress that leapfrogs the American AI companies, despite all the advantages that they have.

[Susan Ettlinger]

So another take that I've seen is the sort of China surveillance and China censorship take, which is like, if this, if this. Or, you know, there's going to be other models that come out very soon like this, as, as you've reported on, as we'll talk about, like, these models will get leapfrogged. This will happen over and over again. But like, at the moment, there's a bunch of people using Deep Seek, it's the top app in the App Store. People are showing that it's censored in the way that you would sort of like, expect Chinese technology to be censored. The obvious example is like, if you ask about Tiananmen Square, it will not answer. And then people also saying, like, hey, this is Chinese propaganda, Chinese surveillance, blah, blah, blah. I'm not sweeping that under the rug. I'm just saying that's an argument that people are talking about. You raised the very good point that ChatGPT is also censored in several ways. Can you just talk a little bit about that and that aspect of the.

[Emmanuel Mayberg]

Yeah, I think the way I think about it is to go back to one of the earliest stories that I did at four four Media, which is about this guy named Eric Hartford who takes these open weight models like Meta's Llama, and because they are open weights and he can modify them, he creates uncensored versions of these AI models. And Joe, I know this always gets to you that they use this terminology, but OpenAI they would consider a censored and woke model. Right? Because it won't tell you how to make a bomb and it won't tell you about the Hunter Biden laptop or so. I don't know if that one's actually true, but it restricts what it says about the election. And something that we bring up all the time is that it kind of refuses to engage with anything remotely sexual things like this. And the argument that the people who make the uncensored models have always made, and this is an argument that Marc Andreessen has made as well, and he has funded Eric Hartford and he has funded these group of small developers who create uncensored models because they argued that, okay, maybe the way that OpenAI censors its models doesn't bother you because you happen to align with their ideology because you live in America and you're American and you don't want the AI to engage with these questions. But if we accept that paradigm, should another country create another dominant AI model and they have their own censorship based on their culture that we don't agree with? We have to live under that. That is the argument for the uncensored model in this moment, this deep seek moment of it showing people that it is possible that we will live in a world where the AI that is dominant is made in China really freaks people out because we then have to accept, you know, Chinese restrictions and Chinese culture and the way that impacts their AI model. Right. It's very similar to the TikTok argument. It's like social media is bad, but we accept it. But the moment it becomes a social media that is dominant as Chinese, people really freak out. And it's the same dynamic here. Facebook does bad things, TikTok does bad things. We are more willing to accept the bad things that Facebook does because it's our social media company as opposed to the Chinese social media company. And you're seeing the same thing here. And I think in a way, though I have a lot of criticisms of the uncensored model argument and community, I think this really proved their point where if you allow for censorship or restrictions on an AI model, just like as a. As a theory of how AI should be managed, then you also have to accept it when it comes from a different country. And that obviously really bothers people.

[Susan Ettlinger]

Right. And it should be mentioned quickly that Deep Seq is an open weights model, meaning you can tweak how it works works and you can run it locally, although we don't have all the information on how it was trained. So it's not fully open source.

[Emmanuel Mayberg]

Yeah, it seems really trivial to be fair to Deepseek or the people who are using it and defending it. It seems really trivial to get around this Tiananmen Square restriction. You just have to run it locally. I've seen people use different languages to get around the restrictions. It's like this is a form of jailbreaking. But I saw someone engage with deep seek in Hebrew and it just answered all the questions. I saw someone get around it by using Leet speak, where you replace letters with numbers. And then it also engaged with the questions.

[Susan Ettlinger]

That's my primary language.

[Joseph]

So I was going to say it's Emmanuel's language.

[Emmanuel Mayberg]

It's a restriction that is there. But the idea that the problem with this particular model of Deep SEQ is that it's restricting information that the Chinese government doesn't want people to have. Sort of, I don't know, unserious because it's open weights and you can get around it. And yeah.

[Joseph]

I think with all of these takes flying. Oh, and I'll just say, I think Perplexity has actually added support for it as well, where you can now interact with this model through Perplexity. And I presume they're running it locally in their own ways, right? Not locally to the user, obviously. So there's all of these takes flying around, but you kind of have one main one in the article and I'm referring to. Even though this is a pretty significant event, is it kind of impossible to tell what's going to happen with the AI industry based on it? What's your main takeaway from this?

[Emmanuel Mayberg]

I think my main takeaway is that the AI space is moving very fast and is developing at the rate of rapid scientific discovery. And that is the schedule that it's on. And that is very hard to predict. We don't know where it's all going to end up. Some people who are kind of cheering and waiting for the AI bubble to burst really hope that this downturn in the market that is instigated by Nvidia and AI companies is going to. This is. This is the AI bubble popping and now it's all going to go away. I'm skeptical of that. Maybe that happens later. I don't know. I don't make predictions. If I was able to make predictions about the stock market, I wouldn't be talking to you, I'd be on my yacht. But so it's like AI is moving at the pace of scientific discovery and the markets are moving emotionally. You know, it's like we're seeing like an emotional reaction in the market. I am, I am doubtful that this was. This will hold. I am doubtful that this is terrible for Nvidia in the long term because they are still the dominant hardware player in this space. It is possible that later on China, especially because we're imposing these export restrictions on them, will develop their own supply chain for the chips. That does devalue Nvidia. But at the moment, Nvidia is the main player and, you know, like one obvious argument about that is, okay, Deep Seq has discovered these new, far more efficient reinforcement learning strategies that allow them to develop models that are better for less. But the research is out there. Nvidia is going to figure this out. Facebook is Going to figure this out. There was a report in a Wall Street Journal, I think that Meta has several war rooms just trying to reverse engineer, recreate this stuff. But the American companies will catch up and then they'll have these new efficiencies and the access to chips, and now we'll just make them better and more efficient. And I also think we talked about leapfrogging. Meta gets ahead, Google gets ahead. These companies kind of vie for the top spot, depending on whose research is latest and best. For a while, they're Mistral, which is this French AI company, was really far ahead for a minute. And people didn't freak out about that. I think because it's a European company and the markets didn't freak out about that. They weren't like, oh my God, OpenAI is doomed. And Meta is doomed because this French company suddenly has a better model. When it's a Chinese company, people, I think, react emotionally because China is perceived as this adversary of the US and the West. Then that's why I think people are freaking out. And I think it's very plausible that a few weeks from now OpenAI will release a new model. Everybody will agree that this is the latest and best, and we'll go back to where we were, which is rah, rah, rah, America and the future controlling AI.

[Joseph]

Yeah, it seems more.

[Susan Ettlinger]

I'm going to go to my war room to determine what this means.

[Joseph]

I think you're in it right now.

[Susan Ettlinger]

It's your Gantt. I'm always in it 24 7.

[Joseph]

All right, we'll leave that there and we'll be right back after the break with another story. We'll be back after this.

[Susan Ettlinger]

It's a new year, so it's time to make some moves. Maybe you start a new business, launch a new product line, or organize your existing operations. For us, this meant a total refresh of all of 404 Media's merch, which we could do only with Shopify. Make this year different and better. With Shopify, the best time to start your business is right now. Shopify makes it easy to create your brand, open for business and get your first sale. It couldn't be easier. Shopify has thousands of customizable templates. No coding or design skills required. All you need to do is drag and drop. Shopify's powerful social media tools let you connect all your channels and create shoppable posts to help you sell your product everywhere people scroll. With Shopify, it's easy to manage your growing business, ship your products, collect taxes and manage payments from a single portal so you can focus on the important stuff, like growing your business. Don't kick yourself when you hear this again in a year because you didn't do anything. Now with Shopify, your first sale is closer than you think. Establish in 2025 has a nice ring to it, doesn't it? Sign up for your $1 per month trial period at shopify.com media all lowercase go to shopify.com media to start selling with Shopify today. Shopify.com media support for the 404 Media podcast comes from Coda. Picture this. You're starting a new project at work and your vision is clear. You know exactly who's doing what and where to find the data you need to do your part. Everything is organized and you don't have to go searching for everything. That's because everything your team needs, from project trackers and OKRs to to documents and spreadsheets, lives in one tab. All in CODA. CODA's collaborative all in One workspace gives you the flexibility of docs, the structure of spreadsheets, the power of applications, and the intelligence of AI in one easy to organize tab. Coda connects tools and teams so that you spend less time switching gears. We're a small team, but we do everything for the whole business. With Coda, we're able to combine what would have once taken lots of different siloed pieces of software and puts them in one place where the four of us can see everything, track our progress, and make moves to help our business. More than 50,000 teams trust Coda to keep them more aligned and focused. To try for yourself, go to Coda IO 404 Media today and get started for free. That's C o D A io404Media to get started for free. Coda io404Media Today's episode is sponsored by.

[Sam]

Better Health in 2025. Maybe you're ready for a plot twist, or maybe there's a part of your story you've been wanting to revise. Life isn't about resolutions that fade by February. It's about picking up the pen and becoming the author of your own journey and therapy. It's like having an editorial partner helping you craft new chapters, set boundaries, and become the best version of yourself. BetterHelp makes getting that support easy. With over 30,000 credentialed therapists and a fully online platform, therapy is affordable, convenient, and designed to fit your schedule. Plus, you can switch therapists anytime at no extra cost. Therapy has certainly helped me find a better version of Myself. And with better help, we've found it's convenient, affordable, and flexible for our busy schedules. Write your story with better help. Visit betterhelp.com 404media today to get 10% off your first month. That's BetterHelp H-E-L-P.com 404media.

[Joseph]

All right, and we are back. And then Sam is now with us. We're going to talk about a story that she wrote. The headline is memos to federal employees were written by people with ties to Project 2025. Metadata shows a lot going on in the headline. Maybe let's just start with what are some of these memos that were sent by the Office of Personnel Management, which for those people who don't know, is basically the US Government's big HR department. That's how I think of it anyway.

[Sam]

Yeah. Honestly, I will admit I had not heard of this office before this week. It's like there are so many little federal offices that I think a lot of people are learning about for the first time because they're in crisis.

[Susan Ettlinger]

Thanks to the only reason anyone would ever know about OPM was it got hacked like seven years ago. And yeah, that's the only reason anyone would have ever heard of this office, I feel.

[Sam]

Yeah, yeah.

[Joseph]

But they're sending these things which are impacting a ton of people. And what are they, like, what are they telling people to do or not do?

[Sam]

I guess, I mean, so they're sending a bunch of memos, which memos are public on their website, so anyone can go and look at them and Download them as PDFs, but they're sent out to federal workers and federal departments. And obviously this is a big thing right now because Trump is, you know, working on firing a lot of federal workers in order to, I don't know, like, so chaos and still people who are loyal to him doing all these kind of like loyalty checks with federal workers and seeing if they, you know, gutting like dei obviously is a big one that's been in the news lately. So these memos go out and the memos say, you know, it's, I can just, I can read you like some of the titles, they're pretty self explanatory. But some of the ones that we looked at just from this week were one was called Guidance on Presidential Memorandum Return to In Person Work, which one of the things that Trump has done is said, you know, all federal workers have to return to the office, whereas before they were allowed to work remotely in a lot of cases. Now everyone who's a federal worker has to be in the office, and that's what that memo is about. And then there's others that are about, like, obviously diversity, equity and inclusion programs. There are, you know, it's just like a, it's a lot of them are, like, kind of dry, but then some of them are pretty, you know, unprecedented slash spicy in general.

[Joseph]

Was a hiring freeze as well, right?

[Sam]

Yeah, there's a hiring freeze. One for, like, federal workers, which is obviously not something that happens every day. I don't know if it's happened, you know, in the last four years, certainly. So, yeah, they're, they're going out to federal workers, basically and saying this is what you need to then disseminate to your department.

[Joseph]

Yeah.

[Sam]

Whatever it is.

[Joseph]

So we'll get to the metadata in a bit and the authors and that sort of thing. But there's been, I would say, a ton of chatter around these memos. And, you know, there's a lot of posting on social media. I would say there's all of these Reddit posts. Some are going to be more believable than others. Of course, you have to take everything a pinch of salt or go and verify it as we do in some cases that we can. And even just beyond these memos, there are multiple journalists tweeting that they've never seen this many leaks, in a way. And I know these memos are public, but it's just sort of associated with the massive reaction that federal employees are having is like a lot of them, it seems, are leaking to other journalists as well. But my question for you is, how are federal employees reacting to these memos? Are you seeing a lot of pissed off people online?

[Sam]

Yeah, I think all of us at this point here at 404, at this point have had folks reach out to us who are federal workers. There's a ton of them. And they're across all different sectors, different industries, working on lots of different things. You know, they're like you said, they're hitting up journalists and saying, this is what I'm seeing. And they're freaked out, obviously. It's like, it's a big deal. They're worried about their jobs. They're being forced to do stuff that they don't agree with. You know, like we'll talk about later, like scrubbing websites and things like that that have to do with like, you know, diversity and inclusion and those sorts of things. So, yeah, they're, they're worried, but at the same time, it's like there's just not a ton that they can do from the inside. They're federal workers. So if they want to like do some kind of protest, it's a federal crime, I mean, I assume. But yeah, it's like they're, they're very disheartened, it seems, about the whole entire situation because they're, you know, they're being told to do stuff that they morally, ethically don't, don't agree with as people.

[Joseph]

So, yeah, I don't think everybody's, I mean, people will be mad understandably, depending on your own circumstances about the need to return to office or something like that. But broadly, many people are simply mad because they don't agree with the incoming administration for various reasons.

[Emmanuel Mayberg]

Yeah.

[Sam]

And like not agreeing and they're like, not like supporting Trump at this point. It seems like that's a fire Wolff, you know, which, that is very unprecedented. That's, you know, it's something that we're seeing, we're going to see more of in the coming days, I think. But like, not like being gung ho about what he's doing seems to be something that puts you on the chopping block as a federal worker now, which is horrifying.

[Joseph]

Yeah.

[Susan Ettlinger]

I think one of the tricky things is that the administration has shut down effectively so many different parts of the government and has also made it very difficult for government employees to talk to outside contractors. And for anyone who's not familiar, like many, many, many, many, like technically third party contractors actually are basically government employees. This is just like how the government works. And it's like there is so much happening because essentially anyone who works for any of these massive agencies can, is being affected in some way, shape or form. And you know, these are literally millions of people. And so trying to tell that story is very, very difficult for us as a small team, but also for like the media in general. And you know, I think you can create like a coherent narrative of what is happening where it's, you know, the Trump administration is just like pausing the government and it's having all these knock on effects. But getting into these specific effects is like essentially impossible because there are so, so many.

[Joseph]

Yeah. The scale is almost unfathomable in a way, and you can't really communicate it in a single article necessarily, or whatever medium you're using. Just because we're talking about opm, I guess we'll touch on this in the subscribers only section as well. But considering opm, they have created these sort of tip lines, right, Jason, where people can email in to, I mean, basically rat on their coworkers when it comes to them running certain DEI programs. Can you just talk about that briefly? And I think you saw that in some of the memos and, and maybe wrote a little bit about it or.

[Susan Ettlinger]

Yeah, I don't have the memo in front of me right now because there's a couple different ones, but.

[Joseph]

And a couple different email addresses.

[Susan Ettlinger]

That's what I mean. I think it's like D D e I a info pm.gov or something. And yeah, dia truth pm.gov where literally they're just asking people to rat on their colleagues who may be trying to smuggle in diversity, equity and or inclusion by any other name, which is really wild and is pretty McCarthy esque. There's also this thing where I believe it's hropm.gov, where it's a sort of master email where the federal government can email every single federal employee, which is not a, like, not a capability that they had before. There was never like a system to email millions and millions and millions of people at once. And so there's been a few tests of this system over the last few days and a lot of federal employees thought that it was initially a phishing exercise because they sent this email from this like, random new email address and asked everyone to respond to it and just say like, hello or like, I got this.

[Joseph]

It's the worst reply all, Fred, in like the history of email. Like people are going to be going, oh, hey, oh, sorry, I hit reply all to millions of federal workers or something.

[Susan Ettlinger]

Yeah, yeah, yeah. But basically it's like this relatively obscure HR agency has just taken it upon itself to become the investigators of the federal government. More or less.

[Joseph]

Yeah. It would be a crazy HR move in a corporate context, we would be reporting on that. Right. Somebody leaks us an email from inside some sort of AI company or tech company or whatever saying, hey, please rat out your co workers who are trying to do certain things around diversity or other programs or whatever. But this time it's the government. I don't know, it's just crazy. All right, let's get back to your story, Sam. So OPM is sending out these memos. They're signed by senior officials, I think, but there's interesting stuff in the metadata. What did this data show exactly?

[Sam]

Yeah. So all of the memos are like you said, they're signed by. Let's see, hang on. It's Charles Ezell, action director at U.S. office of Personnel Management. So it's like that's a very normal person to be sending these memos because he's the director of opm. And then sometimes they have a second author, like a second signature. But that's who they're ostensibly coming from. That's who they want you to think that they're coming from, I guess. But then if you go in and this is something that someone found on Reddit pointed this out, which I don't think many people would ever go to this length to try to look at this metadata. But like you said, it's like people are looking and talking about portions and sections and little dark corners of government correspondence that they haven't before. This person on Reddit said that if you go, if you download the PDF from the OPM website, from the public website and then open it, and I opened it in Adobe Acrobat, but you can open it and then right click and go to the document properties. I think on Mac it's slightly different, but in the document properties you can see the author and the author is someone different. It's someone who's not listed on the front of that letterhead. And that was definitely something worth digging into. And it turns out that the authors that are listed here, that the metadata says at least, are people who've been longtime Trump supporters, longtime like boosters of his agenda, and even in some cases people who helped work on or create systems that set up a very scary agenda called Project 2025, which I think a lot of people have heard of by now. But it's still amazing that some people haven't, considering that's seems to be what we're in. Seems to be what's happening.

[Joseph]

People connected to it have, I mean, I remember straight after the election they were like, okay, actually, yeah, we can come mouth golf now and say that 2025 is the agenda. So there are two names you point to in the piece. Let's just take them one by one. The first piece of metadata says that the, the return to in person work memo, it was signed by that senior official of opm, as you said. But the metadata included the name Peters, Noah. As in Noah Peters. Just briefly, what did you include about that person?

[Sam]

Yeah, so the Noah Peters that, this seems to be his LinkedIn. And this is why I say it seems to be him. His LinkedIn now says he's a senior advisor at the OPM, whatever that means. But you know, it's like his, his background is what's concerning here. It's that he, he is an attorney. He represented a guy named Jared Taylor, who is a really fun Google, but he is, I forget the exact wording but he's, you know, widely known as a white nationalist, and he sued Twitter in 2018 for banning his account and because he's a white nationalist. And he lost that case. But that's. The guy who was representing him is Noah Peters. And then, you know, Peters has been kind of writing and talking publicly about how Kyle Rittenhouse, who killed people at a Black Lives Matter protest, should receive restitution from the media because he's been acquitted. And the media reported on his. His case. It's things like that. It's like he's. Yeah. Just a guy that, like, you don't love to see his name come up in one. In this context, which, you know, he's writing apparently an in person return to work memo.

[Joseph]

Yeah.

[Susan Ettlinger]

Not.

[Joseph]

Not the sort of person you would jump at the chance to ask, could you write a U.S. government memoir to go out to all federal employees? Okay, well, that's the first one. And then the second one is for the memo called Federal Civilian Hiring Freeze Guidance. What was the name on that document, according to the metadata?

[Sam]

Yeah, so the metadata says that that one was authored by James Shirk. And Shirk is also a fun Google. Just guys that I wish I didn't have to know the names of. He was a special assistant on domestic policy in Trump's first term, and now he's back. But in the meantime, he was working at the Heritage foundation, which is the think tank conservative group that architected the Project 2025 playbook. And then he was also on the Trump transition project called America First Policy Instit. And what's really important here and why it's important that he apparently seemingly wrote a memo called Federal Civilian Hiring Freeze Guidelines, is that he came up with the classification that makes it easier to fire federal workers. That's called Schedule F. That removes their employment protections, makes it easier to get rid of them. And that is mentioned several times as a really key part of Project 2025, because Project 2025's playbook, which is a long. It's like 900 pages or something. A long book.

[Joseph]

Yeah. I ain't reading all that.

[Susan Ettlinger]

Yeah.

[Sam]

And. Yeah, well, you can go and read it on your free time, Lister. No, no, we won't do a dramatic reading. Yeah. But yeah, it's like, mentioned like Schedule F, which is the. The classification that Shirk invented, comes up often because it's this kind of, this real linchpin to get rid of a bunch of federal workers to then reinstall people who are loyal to Trump. And that's a Huge part of Project 2025's entire mission. Without that, it's really hard for them to get any of the other stuff that they want to do done. They need to overhaul the federal government in order to push forward a lot of these plans. And a lot of their plans are things like restricting reproductive access, mass deportation, and again, firing, you know, tons of civil servants, and reinstalling, you know, like, these people who are ride or die Trump. So, yeah, he's a big one.

[Joseph]

And I think those are the two you mentioned, right?

[Sam]

Those are the two. Yeah, they're. And, you know, if you go through and you can. People can look at this themselves. I've seen a couple people say online that the documents are now. The metadata is scrubbed. That's not what I found. I was still able to see the author data when I checked earlier today. But you could go through and do this yourself. You can look at it yourself and see the different authors and see that they're not. Who is on the front of that actual letterhead. And that would be a normal thing. If you have someone writing your memo for you because you're the director of opm, that's. That seems. I was like, okay, fine. But, like, these are not just, like, guys who report to the director. Like, if anything, the director is probably, like, taking calls from them.

[Joseph]

Right.

[Sam]

I would imagine. You know, I. Obviously, we don't know a lot of this stuff for sure, but, yeah, they're not just, like, average, like, administrative assistants. They're. They're people who wrote the playbook for fascism. You know, that's. That's what we're working with.

[Joseph]

They're not a secretary or assistant. Right. So Sam touched on this, and I think, just to sort of ask you, Jason, a slightly different way, why is it important to know who the people who seemingly wrote these memos are? Why does that matter?

[Susan Ettlinger]

Yeah, I mean, I think it's important because Trump swore up and down that Project 2025, he had nothing to do. Do with, he knew nothing about, was not going to be involved with, blah, blah, blah. You know, that that is what, like, the Democrats tried to tie him to project 2025, and rightly so, because this was obviously the playbook from the beginning. It was the transition plan from the beginning. But this is, like, hard evidence showing that these people are immediately in the government, are immediately writing, like, incredibly consequential memos, and are immediately having an impact on government. You know, unclear whether, like any. Anything resembling normal politics matters now or will ever matter again. But it's Just evidence that, yes, like, Project 2025 was the plan. The whole time is the plan. Now these people are in the government, are enacting policy, are making decisions, are writing memos that have, you know, real impacts on people's lives and how people interact with the government. And to the extent that, like, anyone gives a. About any of this stuff anymore, which I think that they should, you know, theoretically, that matters.

[Sam]

Yeah. And Trump distanced himself from Project 2025 this whole time when he was running because it was poisonous. Because most, most people hear, you know, the things that were laid out in it and they're like, oh, that's, you know, Christian nationalism. We don't, like, we're not into that. That's like. And he distanced himself from it on purpose. He was like, well, that's, that's not really what I'm into. I'm just, you know, for the economy or whatever the. But, you know, Biden was like, well, what about Project 2025? And then just kind of let it go. And then it came out that, you know, Trump immediately started hiring, like over 100 Project 2025 architects into his administration. So go for it.

[Susan Ettlinger]

I would like to throw a curveball at all of you at the end of the show, like I do often, which is while we've been talking, the White House did a press conference about the New Jersey drones that we have been talking about multiple times on the show in the past. And the White House press secretary said, quote, this was not the enemy and they were authorized to be flown by the faa, the, you know, drones that were being sighted above New Jersey for like weeks toward the end of last year. And that I reported on, wrote and Frant and Ray had done this podcast about how this happens all the time. It's a media caused panic. You know, talk of conspiracy theory and aliens and all that is, like, very dumb because we've seen it before. And I think it will be very, very, very interesting to see how the media covers this, how social media covers it, how like right wing influencers talk about it, et cetera, because the Biden administration said the exact same thing. They said these are FAA hobbyist drones and. Or like, you know, we know what they are more or less, and it's no big deal. And when Joe Biden said it, or Joe Biden's press secretary said it, it became the biggest deal. And it was on every single cable news channel for weeks and everyone was talking about it endlessly. And now here, here's Trump and Trump's administration saying the exact same thing. And everyone is going to fall in line immediately. The media is going to do a story saying, oh, it was no big deal. They're never going to talk about it again. And it's like, that is a really big problem. It's a really, really, really, really big problem. Because when the government, like, under a Democrat says, hey, like, here's what's going on. And the response from half of the nation is, this is bullshit and it's a conspiracy. And, like, up and down across, you know, the sort of cable news ecosystem, you pull on all of these, like, frankly, whack jobs to shriek about something for a while, and it becomes this, like, entertainment value proposition. Like, that is not. That is not a media, media ecosystem that we can abide. And it's like, it's really. It's a really bad problem. And it is related to what we're talking about here because it was clear to everyone that Project 2025 was going to be the plan forever because that the Republicans spent years and years writing this down. A thousand pages, putting people in positions of power and in positions to enact this, like, on day one. And Trump was able to say, like, oh, actually, no. And it went away for a large segment of the population. And it's just like that. That capture of our, like, information ecosystem is very, very bad. And I'm sorry, we. This is not our podcast document, but, like, no one is going to care about the drones ever again. It's crazy that this is happening.

[Joseph]

One clarifying question, and maybe you don't know, but since you put a curveball on us, I'm going to strategically bat it back. I don't know. I don't watch baseball. But the quote from this press conference is that after research and study, the drones that were flying over New Jersey in large numbers were authorized to be flown by the FAA for research and various other reasons. So there were government drones, they were being flown by the government. Or, like, how are we supposed to read into that? Or how do you think, Griffith?

[Susan Ettlinger]

So that's a great question, because getting authorized or flying by getting authorized. Getting authorized by the FAA does not mean that they're being flown by the government. It means that they have, like, clearance from the faa. So it could be university drones, it could be hobbyist drones. It's like, you have to get a license to fly a drone now, if you're flying for specific reasons. So it could mean that. And then the rest of this quote, which you tried to gotcha me. I did. I deliberately left it out said many of these drones were also hobbyists, recreational and private individuals that enjoy flying drones, which is sort of like what we were saying when we were doing our reporting on it, because a lot of people are flying drones for like whatever reason. And there's not. Like, you do have to get an FAA license in some cases, but they then don't control where you fly. And so.

[Joseph]

But the way I'm reading that second part, that many of these drones are also hobbyists. Again, maybe I'm reading too between the lines here, but to me that sounds like, well, there were the FAA authorized ones and then people started flooding the airspace with their garbage drones, all trying to jump on it, basically. Which is kind of what you said earlier, right? It's like.

[Susan Ettlinger]

Yeah, so, I mean, I'm curious. Like, you know, there were a bunch of drones being seen and there's things being flown all the time. So it could be like it. You could. Sometimes you need specific FAA authorization to fly a specific flight path if it's near an airport, for example. You know, the Trump administration has not released these documents about, about like what the authorizations are, so it's unclear. Yeah, it could be. They could be government drones to some extent. I, I believe there's another quote in here where Trump was saying something at some point. Quote, look, our military knows where they took off from. If it's a garage, they can go right into that garage. Not sure what that means, but good point, good point. Yeah, Yeah. I would guess that it was a mix of military drones, perhaps like NOAA or NASA drones, maybe like university research type stuff, and then also hobbyists and, you know, I mean, maybe this will continue to be some big conspiracy. I know that this was a month ago and people haven't been talking about it since the sightings magically went away around Christmas when people were like paying attention to something else and there wasn't coverage.

[Joseph]

Exactly.

[Susan Ettlinger]

I know there wasn't coverage, but, you know, I. There's a Guardian article that we're talking about. Like, I wonder if this is gonna. I mean, it's not going to. Because it shouldn't, like lead CNN and Fox News and all that for hours and hours and hours upon end. Like, kind of doubt it.

[Joseph]

Yeah, fair. All right, we'll leave that there, but I'm sure, I mean, Jason, you should probably go write about the drones now. Probably for tomorrow or something.

[Sam]

Yeah, you just dictated a blog, so please go get that.

[Susan Ettlinger]

We'll do a text to speech.

[Joseph]

Yeah, well, until then, we will leave that there. If you're listening to the free version of the podcast, I'll now play us out. But if you are a paying 404 media subscriber, we're going to talk about how GitHub is showing how the administration is scrubbing material from its websites and across the web in real time. You can subscribe and gain access to that content at 404 Media co. As a reminder, 404 Media is journalist founded and supported by subscribers. If you do wish to subscribe to 404 Media and directly support our work, please go to 404 Media co. You'll get unlimited access to our articles and an ad free version of this podcast. You also get to listen to the subscribers only section where we talk about a bonus story each week. This podcast is made in partnership with Kaleidoscope. Another way to support us is by leaving a five star rating and review for the podcast. Somebody did that recently. I was going to read it out and I forgot to put it in the document. So do that next time. This is been 404 Media. We will see you again next week.