Google Employees Meme About How Bad Their AI Is

0:00I was just listening to all these CEOs talk about how much of their code is being produced by AI. That just doesn't line up. The idea that 75 % of the code is the most influential tech infrastructure company in the world is being AI generated. That's not the complete picture.

0:21Hello, 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 404media.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 404media.co. I'm your host, Joseph. And with me are 404 Media co-founders, Sam Cole. Hey. Manuel Mayberg. Hello. And Jason Kebler. Hi. All right, let's get to our first story, which is one that Emmanuel just published.

1:03The headline is, Google employees internally share memes about how its AI sucks. So you've been sent a lot of these memes recently, Emmanuel. We will display these on the YouTube version of the podcast. If you're listening and you'd rather have video, you can always check out the YouTube version. There's a link in the show notes. So you just go to our YouTube channel and obviously subscribe. And then every week, you're going to get the video version of this podcast. But for those listening and to run us through it, can you just give us some examples of these memes, first of all, before we drill down into them?

1:41Sure, yeah. So I really think that the meme I put at the top of the story sums up the sentiment of employees at Google and how they feel about their AI tools, specifically JetSki, which is Google's internal AI coding assistant. And this meme is the doubt face, press X to doubt face from LA Noir. Is everybody familiar? I think we should do a survey as we know. I didn't super know this meme, honestly. I got what it was going for, but I don't know this guy. Press X to doubt. It's like press F to pay respects. That's right. In the game, you're an investigator and you're interrogating different people and you can react to their statements in various ways.

2:42And in this one, the main character is making a very skeptical face that he can press X to doubt. And the impact font says, Jetski built a functional operating system autonomously, which is something that Google recently claimed at its annual IO conference, which it's an annual conference. They've been there for many years. They just talk about all their biggest upcoming stuff. And understandably, this year, it's been heavily focused on AI. They showed off a bunch of AI products, and they made a bunch of outlandish claims like their AI was able to build an operating system from scratch. A lot of the memes are about JetSki, but a lot of them are also in response to stuff that they saw at IEO.

3:27So I think that's a really good contrast, which is what the story is about, essentially, and the line of reporting I've been interested in in the past few weeks, which is what is Google saying publicly, what other companies are saying publicly about their AI products, and what the people who actually make them and use them say about them internally. Yeah, I'll just give one more example. There's the big forehead fish. Do people know this meme before we get into it? I know that one, yeah. The diver. It's originally a video of a diver doing some meticulous work on, I don't know what, is it a shipwreck or coral or something?

4:04And a fish, I'm sorry, I don't know the name of a fish. It's probably not even a fish, it's some sort of sea mammal, I don't know. It's a dolphin with a giant head. It's not a dolphin. Giant head dolphin is the name. And he's coming up to the diver and he's like, what are you doing? What are you doing? He's like bugging him. And then in this internal Google AI version, the fish is like, aren't you using AI? Why is it still taking so much time? AI is magic. Are you a muggle? New best AI tool launched just today. And the scuba diver is just trying to get on with their job. Like that is the sentiment that seems to be coming out from these memes that isn't...

4:41Well, sometimes, as you say, it's about specific products like JetSki, this internal tool or whatever. And then it's like, we're being pressured to use these AI tools, even when we may not want to, which kind of reflects on the earlier stuff you had about the leaderboard at Amazon, right? And this is just like another way that employees are sort of saying how they feel about this, right? Is that sort of the vibe you got? Yeah, I actually wanted to talk a little bit about how I got into these stories, which is sometimes we get stories because people send us tips. Sometimes it's because we see something in the news.

5:20And sometimes it's just because we have a feeling that something is a story. And in this case, I was just listening to all these CEOs talk about how much of their code internally now is being produced by AI. 20%, 50%. Google famously, as I mentioned in the lead of the story, more than 75 % of the code at the company is being produced by AI. And judging by everything we know about these AI tools and how often they make mistakes, this is something we've covered in many ways and also specifically about AI coding agents. It's like that just doesn't line up. Unless Google has some super secret, super powerful AI that doesn't make mistakes like every other LLM does.

6:05It just like the idea that 75 % of the code at like one of the biggest tech companies, one of the most influential tech infrastructure company in the world is being AI generated. It's like, surely there's like a different, that's not the complete picture, right? and we strategically published a story that was based on a lot of conversation that was happening on Reddit that showed exactly this, that people were complaining about how they're being pressured to use these AI tools to produce code, whether it was useful or not. We kind of did a broad story about that sentiment and asked for tips. And I would say, this is a strategy that we've used forever.

6:49We started using it in a motherboard. it's very effective. I've never seen it be this effective before in my career. I published the story and I've heard from dozens of employees across the industry that were dying to talk about this problem. And yeah, I mean, to your question, it's like, this is the sentiment at Google and this is the sentiment at other companies. Some people feel like the AI tools are useful. Some people feel like they need more work and they'll get there eventually. There's a broad spectrum of what the sentiment is about AI tools from very positive to completely apocalyptically negative.

7:32But everybody agrees that they're being heavily pressured to use AI, that that pressure is... There's no other goal other than to just show that they use AI. It's not based on outcomes. It's not based on increasing productivity. It's not based on anything like that. Just like we want to show that we're using AI, whether it's helpful or not, and that it's driving the employees crazy in many cases. Yeah. And they're not just making these memes, but then people internally at Google are essentially upvoting them, right? Or reacting to them. Some of these memes have hundreds and hundreds of likes or whatever.

8:13So it's not just the people making the memes. Everybody's like, yeah, this fucking sucks as well. I mean, just one of the... Maybe people don't know this, but these tech companies have internal message boards. We've seen stuff from the past from Facebook and I think other tech companies over the years. Just briefly, what is this message board for Googlers? Is it like a forum? Is it just for memes? What's the deal? Yeah, so I think this is pretty weird and funny, but Google has many ways of communicating internally. They have Slack and different forms and company registries and all that. But they also have something called MemeGen, which is like an internal social media just for producing memes.

9:03And this has been in the news before because, in a way, it's like a temperature check on what employees across this giant company are thinking about and talking about and joking about that you can see what they're memeing about. So I think the New York Times had a story maybe late last year about how they actually started to enforce some rules around the memes that people were making about conversations around Gaza. So that's been in the news. So it's like management definitely keeps an eye on it. In terms of AI, what I've heard is that it's like a third or a quarter of all memes that the company are about this subject now and that it spikes whenever Google talks about it publicly like they did in IO or when something breaks down internally, right?

9:58Like everybody's trying to use JetSki, JetSki breaks and then everybody rushes to Memegen to make memes about it. So it's like a dominant sentiment at the company, or at least it's one way to measure it. Yeah, it's super interesting, in my opinion. So Google gave you this statement that I'm just going to read out. We encourage our engineers to vigorously test and critique our internal tools. That candid feedback loop is vital to how we build technology. AI coding models are designed to assist developers, But it's critical that we maintain humans in the loop, including the oversight and expertise of our world-class engineering talent.

10:40We continue to refine our internal tools based on employee feedback to ensure they are enabling and not hindering daily productivity. Okay, sure, whatever. You get that statement, Emmanuel. We then publish the piece. And I think you get an email just before or just after, but Google wants to change their statement. Can you explain what the deal with that is? Yeah, so I published a story with their statement. And then right after I sent out the story in our newsletter and published it on our website, I get an email from the same PR person. He was like, actually, do we have time to update that statement?

11:23And I said, I just published. But obviously, if you would like us to update the statement or add another statement, we could do that. If there's more interesting information for our readers to know, that sort of thing maybe they legitimately missed out in important detail or something. Maybe. Yeah. And then he sent the updated statement. Do you want to read that one? I don't have it in front of me. Yeah. We encourage our engineers to vigorously test and critique our internal tools. That candid feedback loop even via our internal meme generator is vital to how we build technology. We continue to refine our internal tools based on employee feedback to ensure we are delivering the best experience that maximizes daily productivity.

12:11So I don't know how you would... What do you read in that edit? I'm not sure. But clearly a lot of eyes on this. I mean, the thing that stood out to me was that they have explicitly removed the bit that says it's critical we maintain humans in the loop. I was like, that was actually a pretty interesting thing to include in your statement. Like saying, hey, we are bullish on AI, but it's very important that humans are still involved. Wait, scrap that. We don't want that in the statement anymore. That's even more telling, in my opinion. What do you think it means? Because I have a take on that as well.

12:47But what is that signaling? I think you'll have a more informed take than me. but for me, it's probably or could be like just a PR move in that, wait, we don't want to acknowledge that we probably should have humans in the loop when we're talking about 75 % of code in Google is now AI generating and stuff. It's like, no, no, no, we're all in on AI and we don't want to give even the perception to, I don't know, investors, whoever, the market, I don't know, that we're not that. So for me, it was that sort of positioning. But what do you think? No, I think that's spot on. I will scoop myself on the pod because I haven't decided if I want to publish a story about this yet.

13:33But another way that I've seen that same messaging expressed internally at Google is there is a document that software engineers at the company can refer to for expectations, like soft expectations. It's not something that says, you have to work this way and this is how you write code. But it's like, here are best practices, right? And there was a message that was sent out in the last few days that edited that document, which is already filled with a bunch of guidance about AI use. And the note about the update highlighted the fact that AI coding or AI-assisted coding is now just referred to as coding.

14:25So it's like to remove the distinction between somebody writing code and somebody writing code with an AI agent is not a distinction they want to make anymore, which I think is sort of the same idea. You want to show that it's like a completely normal, legitimate way that the company works now. And that may not include a human in the future. That is really interesting. We'll leave that there. When we come back after the break, we're going to talk about Jason's story on Microsoft basically saying the quiet part out loud. Wait, wait, wait, wait, wait, wait, wait, wait, wait, wait, wait, wait. Breaking news.

15:09Is it about Lego? It's the Asian sheephead wrasse is the type of fish in the meme. Is it a fish? It's a fish, not a dolphin. It's native to the Western Pacific Ocean. It lives near Korea and China and Japan. Respect for the fish. Yeah, and it's super crazy because it looks like a normal fish when it's a kid, but then as an adult, it grows this big bulb on its head. Just like me for real? Yeah. All right, we'll be right back after this.

15:50This message is sponsored by Raycon. I think the hardest people to shop for are dads. Not because they don't like gifts, but because most of them don't actually need more stuff. Another mug, another gadget that ends up in a drawer? Probably not. That's why I like gifts that become part of someone's everyday routine. And Raycon's Everyday Earbuds Classic are exactly that. Whether he's listening to music while working around the house, taking calls during the day, watching videos on a tablet, or catching up on podcasts, these are the kind of earbuds that get used constantly. I feel like I always have them in my ear.

16:25One of my favorite features is the active noise cancellation. Sometimes you just want to tune out the background noise and focus on what you're listening to. And these do a fantastic job of making everything sound more immersive. They also come with multi-point connectivity, so you can pair them with two devices at once, plus a super comfortable ergonomic design that stays put all day. And the new colors are great too. Personally, I'm into the cool mint version, but I also have a few other colors too. It stands out without being over the top. The best part, Raycon delivers premium audio quality at about half the price of the big brands.

17:00Add in more than 3 million happy customers and a 30-day happiness guarantee, and it's an easy win for Father's Day. I've been wearing my Raycons pretty much every day and everywhere I go. whether that's at the gym, running errands, working from home, walking my dog. I even take calls on them like when I'm walking around, work calls, stuff like that. So I've made them part of my everyday routine and you can too. Upgrade your dad's everyday routine. Go to buyraycon.com slash 404 and get 15 % off. Thanks Raycon for sponsoring. Every summer I end up doing pretty much the same thing. I stop thinking about outfits and start thinking about comfort.

17:39When it's hot out, I basically want a simple uniform. A great shirt, comfortable pants, and fabrics that don't feel like they're fighting against the weather. That's why I've been reaching for Quince lately. Quince makes high-quality essentials that feel way more expensive than they actually are. Their European linen shirts, which is exactly what I have right here, and they have pants that are linen too, have become a staple for me because they're lightweight, breathable, and somehow still look put together even when it's 90 degrees outside. Honestly, they've become a bit of a uniform for me. I feel like I'm wearing the linen pants all the time.

18:09But it's not just linen. Their tees are incredibly soft and their lightweight cotton sweaters are perfect for chilly nights, especially if you live by the beach. What's really surprising with Quince is the value. Everything at Quince is priced 50 % to 80 % less than similar brands because they work directly with ethical factories and cut out the middlemen. You're paying for quality, not giant brand markups. Honestly, I have become pretty obsessed with Quince. I now have a Quince rug. I have Quince sheets. I have a Quince duvet cover. I have a Quince pillow. And then again, I have Quince shoes.

18:41I'm wearing a Quince shirt right now. So bedding, cookware, furniture. It's one of those brands where you keep finding things you didn't know you needed. And because it's Quince, you know it's good quality. Elevate your summer wardrobe. Go to quince.com slash 404media for free shipping on your order and 365 day returns. Now available in Canada too. That's quince.com slash 404media for free shipping and 365 day returns. quince.com slash 404media. At 404media, we know your personal information is everywhere online. Data brokers crawl the web to build profiles, selling your home address and phone number to anyone with a credit card.

19:20Trust us, we know this better than anyone else. This exposure leads to risks like doxing and identity theft. That's why we recommend Delete Me, a hands-free service that removes your personal info from hundreds of data broker sites. Their experts find and remove your information, providing a detailed privacy report within seven days. Delete.me monitors these sites all year long to ensure your data stays removed. We feel like we're at increased risk because of the types of stories we cover and the public nature of our job. But the truth is, identity theft, threats, and harassment can happen to anyone.

19:55I've been In fact, lots of us have been doxxed. And Delete.me has helped us pick up the pieces, get our data removed from shady websites, and otherwise clean up our online presence. I've been using Delete.me for years. And every quarter, I look forward to seeing what new, interesting, and horrifying websites they've had my data taken down from. It's kind of a perverse little game. Take control of your data and keep your private life private by signing up for Delete.me. Now at a special discount for our listeners. Get 20 % off your Delete Me plan when you go to joindeliteme.com slash 404media and use promo code 404media at checkout.

20:29The only way to get 20 % off is to go to joindeliteme.com slash 404media and enter code 404media at checkout. That's joindeliteme.com slash 404media, code 404media.

20:52All right, and we are back. The headline of this one by Jason is Microsoft wants... Sorry, it's double bylined, isn't it? It was originally Jason, then Emmanuel wrote as well. I'm going to acknowledge that. Apologies. Written by both of you. Microsoft wants to, quote, make people addicted, end quote, to its new AI assistant internal documents reveal. So, Jason, just tell us about this addiction line first. We'll get into the different phases and stuff, but this is obviously the most important part. What is this internal document and what's it saying about addiction? Yeah. So it is an internal document about a new product that when I was writing this article had not been announced or released yet.

21:40And it was called Claw Pilot. And it was basically Microsoft's attempt to use OpenClaw on its Microsoft 365 products. We can talk about what all that means in a minute. But basically, it was a strategy document for this upcoming product. And it had three phases for the deployment of this product and what the plan was. And phase one was, quote, make people addicted. and that is pretty crazy to see in an official company document because AI companies have been repeatedly sued for making an addicting product. They've been repeatedly attacked by... I mean, correctly so, but basically it's like that word, addicted and addiction is like a third rail for social media for internet companies for for artificial intelligence that like politicians don't like that regulators don't like that courts don't like like that sort of thing and here here it was as them putting it as the very first thing in this strategy document for this new product yeah you're taking notes on the criminal conspiracy and you're putting it in writing i like straight up i think this is going going to lead to some sort of legal action, surely.

23:07The fact that they just came out and said it. But we'll talk about that in a minute. So they're preparing to... It's definitely like a they admit it situation. Right. So eventually then they actually launched this product and then we publish... Or sorry, they announced this product and we publish on the same day. And that's called Scout now, right? Before we get into more about the document, just what is this product then? Like it's one of these new agentic AIs that goes and does stuff for you. And this is like Microsoft stab at it? Yeah. So basically, for a while, Silicon Valley was obsessed with OpenClaw, which is an open source tool that allows you to create AI agents, which are basically like...

23:55It's AI that does stuff. So it's AI that can send emails. It's AI that can add things to your calendar. It's AI that can act on your behalf. And to do this, you need to give it access to a bunch of your accounts or you need to make accounts for it. So you either need to have a standalone online world for your AI agent where they have their own email address, they have their own calendar, they have their own, in some cases, bank account. The things that people were doing with OpenClaw were quite wild and led to all sorts of crazy things happening where they were making phone calls. And I say they because I don't know what else to call them.

24:36They're not people, but I'm just using that terminology not to anthropomorphize these agents. But basically, this went super viral and then every big AI company was like, oh, we need some version of this because right now we're basically stuck in chatbot land where you talk to a chatbot and you have to then copy-paste the email that it writes for you over into Outlook and send it. And so this is basically Microsoft's attempt to take OpenClaw and to integrate it into their products. And so Scout is what they're ultimately calling this. It is an agentic AI that will live on Microsoft 365, meaning that yes, it can add to your calendar, it can send emails, it can...

25:31Maybe it can delete files. It can like... In OneDrive or something, presumably. Yeah, it can interact with all the Microsoft products that you have and then also with Windows more broadly. I mean, that's when it would start to get a little bit crazy and you're going to end up like that meta employee who deleted their entire inbox or whatever. So that's the idea. That's the product. As you already said, this is what Microsoft wants to make people addicted to. The actual quote is make people addicted, right? And they want to do that with this. What is the idea after that? Because this document lays out a couple of phases, right?

26:14Yeah, so I think it's important to say that the document itself was written by AI. So the person who is doing this is a Microsoft executive named Omar Shaheen. And as he describes it, he's now written many blog posts and LinkedIn posts about this. And all of them have been AI generated for what it's worth. Like very obviously AI generated. You can just tell. One, you can tell, but then also Pangram, which is the tool that we've mentioned a few times before that detects AI writing. is like, this is 100 % AI. And so that can be not necessarily foolproof, but they are AI generated for sure. Because it reads that way, but then also the actual document that we obtained says that it was generated by AI.

27:11It's like, let me see exactly what it says. Like assisted? It says, co-create the document that we obtained or that we saw and wrote about. At the top, it has Omar Shaheen's name on it. But then it has a subheading and it says, quote, co-created turn by turn with AI, human verified every sentence. Including make people addicted. Including make people addicted. Yeah. And so I just say that because phases two and three are like inscrutable to me. they don't really make very much sense. They're kind of boring. Then they're kind of boring. So phase one is make people addicted. Continue shipping the standalone Claw Pilot experience.

27:56Polish the UX to grow the user base and build the skill and tool ecosystem that makes people depend on it daily. This is already happening organically. Phase two is connect with Lobster. Lobster is like the broader Microsoft plan to integrate OpenClaw in its products and have agentic AI, more or less. And then phase three is agentic identity and new significant planned features. It says, once connected, expand into mobile app for voice, governed skills and MCP marketplace, agent-to-agent collaboration, co-creation tools for team-level skill authoring, and deeper MCP server ecosystem integration with the graph team's enterprise MCP offering.