The 404 Media Podcast (Premium Feed)

Joseph: 00:00

This isn't about safety. This isn't about alerting the FBI if somebody is about to do a really serious crime, they're banging on ChatGPT. This is just to make the models better and to make the responses from ChatGPT better in OpenAI's eyes. So, you know, presumably more people will use the product. Hello, and welcome to the four zero four Media Podcast where we bring you unparalleled access hidden worlds both online and IRL.

Joseph: 00:29

Four zero four 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. Get access to that content at 404media.co. Also, do remember to subscribe to our YouTube channel where you can watch all of our episodes.

Joseph: 00:55

Subscribe at youtube.com/404mediaco. I'm your host, Joseph, and with me are two of the other four zero four Media co founders. The first being Sam Cole.

Sam: 01:08

Hey.

Joseph: 01:08

And the other being Emmanuel Babo. Did you forget my name? No. I was trying to cough at the same time. I was like, do I tactically turn my voice turn my mouth away and try to sneak it in?

Sam: 01:22

Emmanuel Baber.

Joseph: 01:23

I actually thought I pulled it off. Emmanuel Baber. No, but you're here as well. Actually, Emmanuel, we didn't prep this, I wanted to ask just because I mentioned the YouTube channel, we actually uploaded like a thirty minute discussion between you and Jason, right, about all of this AI, doom, AI is gonna kill us, there's a 10% chance it ends humanity talk, right? Just briefly, like what did you two talk about in that news?

Emanuel: 01:54

Yeah. So it's a new thing that we invented that day that we're going to try. I had a lot of fun doing it. Basically, Jason and I, and I guess the rest of Slack, we we we just spent all week talking about this thing that was very viral on social media. This researcher left Anthropic said he couldn't ethically be part of the company because the company was very irresponsible and that AI would be a danger to humanity given how they're developing it, then someone who also works at the company and stayed said, like, that's true.

Emanuel: 02:29

We did it. We do think it could end humanity, but we're trying to do a good job. And that started a media storm that is ongoing this morning. I was listening to a news podcast about that, and we had a lot of thoughts about it, but not something that was good for, like, a discrete concrete article. So Jason and I just turned on the cameras and chatted about it for like half an hour, and I think we had a pretty good discussion.

Emanuel: 02:55

If you want to know what we think about all of that, even though there's no story about it on the website, please check it out. Hopefully we do it again kind of not not once a week, but whenever there's something like really big in the news that we have a lot of thoughts about, but not something that we want to write an article about.

Joseph: 03:13

Yeah. It's almost like an emergency podcast where it's less produced. We can just get on Riverside, the platform we use, and then just talk a bit and then throw it out on YouTube with a thumbnail and that sort of thing. I do think we should explore that more. But I wrote this first story.

Joseph: 03:33

Emmanuel, do you want to take the lead on this one?

Emanuel: 03:35

Yeah, so the headline for this is Inside Project Lily, the Humans Reading Your ChatGPT Chats. So let's start to get our head around what Project Lilly is. Do you want to do that by just maybe reading the lead?

Joseph: 03:54

Yeah. I think that's probably the best way to summarize it. The gist is that OpenAI, obviously the makers of ChatGPT, are hiring, you know, hundreds or at least hundreds of contractors who read this massive stream of real users ChatGPT prompts, and sometimes those prompts, some of which I've seen and we'll get into that, include, you know, pretty sensitive personal information. Now, to me, what really makes this stand out, and I'm sure we'll talk about all the different ways that humans have been involved in AI training. We know that here, humans are looking at real ChatGPT user prompts.

Joseph: 04:36

Like this is actual user data that these people are looking at, which I think is pretty important when people increasingly use ChatGPT or LLMs more generally as, you know, a therapist, even though they probably shouldn't, but they use it as a therapist, they use it as a digital friend, they use it for all sorts of business tasks or whatever. They might even ask it sensitive questions that they don't want to ask another person. And in fact, lot of OpenAI's marketing at the moment, I don't know if you've seen this, but like maybe not TV ads, but YouTube ads, that sort of thing. It's like, hey, when you have a question that you don't want to ask a person, you can ask ChatGPT and we won't judge and we'll just give you the answers. Well, turns out there might be a contractor reading those conversations as well.

Emanuel: 05:27

So how do we know this? What materials did you get your hands on that informed us that this was happening?

Joseph: 05:34

Yeah. So I've seen internal documents that lay out this process that includes instruction guides for the contractors who are reading these ChatGPT prompts and conversations. It includes material from Slack channels as well. I've also seen the specific dashboard that these reviewers are using where they log in and then they are presented with the prompts and sort of the tasks that they have to do. So I've seen the actual back end of that as well.

Joseph: 06:09

And, you know, most importantly, I've seen multiple actual ChatGPT user prompts from that system as well. So I've seen basically the system that the contractors are using to read ChatGPT prompts. I've seen the prompts themselves and like the training materials that these contractors have seen as well. So, you know, it's a pretty in-depth cache of information, I would say. And then I spoke to somebody who was working on these prompts as well, and they they provided more context.

Emanuel: 06:42

So the story doesn't repeat verbatim the chats we saw because we want to respect users privacy. But can And protect the source as well. And protect the source. So OpenAI doesn't use those specific chats to try and locate who reviewed them. But what can you

Joseph: 07:04

tell us about what ChatGPT user chats may be read by humans? Yeah. So it is really, really varied. When a contractor logs in, they will see the prompt. And sometimes this is very, very short.

Joseph: 07:24

Sometimes it is a bit longer and more developed like a paragraph. Sometimes it is a full on conversation between the ChatGPT user and the chatbot itself. Very interestingly, sometimes at the top it includes like ChatGPT's memory. So we will say stuff like, hey, this user is located in this part of America and they previously have used ChatGPT to do X, Y, Z and that sort of thing. So there is like a lot of information about the user in there.

Joseph: 08:04

I guess I should just say this briefly that OpenAI says it uses a privacy focused model to process these prompts before they get to the contractors. And the idea is that it will remove personal identifying information and that sort of thing, so it's not exposed to the contractors. I'll say two things on that. The first is that OpenAI acknowledges it may not always work because the contractors are told, hey, if you see personal information, please escalate it. So clearly, this privacy model doesn't always work and maybe they're still trying to improve it.

Joseph: 08:40

Second of all, I would say, based on the prompts I've seen, I think there can still be enough context in the prompt itself, which may not be literally personally identifiable information like somebody's phone number, where you could reasonably figure out who this person is or give it like a very, very good shot. So there's sensitive stuff in here. And although some of the prompts I've seen as well, they're just like very banal and like, frankly, quite bad. If people are giving these prompts to a ChatGPT, they're probably not going to get the best answer out. Some of them are very, very casual, but some of them are very, very detailed as well, which I think is, you know, obvious when ChatGPT has something like 900,000,000 users, right?

Joseph: 09:29

There's going to be a really wide spectrum, sort of the quality of these prompts. But regardless, the point is that some human somewhere, there is a good chance or sorry, there is a chance that a human is going to see these.

Emanuel: 09:41

Sorry, this is a bit of a tangent, but I was thinking about this when I was reading the story. This thing about them having a model that strips the chats of personally identifying information is very interesting and like a new concept as far as I know. Like, you have done a ton of reporting on how content data that companies claim is anonymized can be easily used to de anonymize anyone. It could be de anonymized and locate specific people, and that's usually by the context, right? You can strip out a name or an address, but if you know who you're looking for, or depending on what the data is, you can just figure out who it is.

Emanuel: 10:26

And we know that is true about anonymizing data previously. I just wonder how this model works. Like, what is it trained on? What are the instructions? Like, is it just removing names and addresses and job titles?

Emanuel: 10:42

Or is it somehow trying to like alter all the contextual information? We have no idea. It's just like a very strange concept.

Joseph: 10:52

Yeah. It is very, very interesting, and OpenAI pointed me to a post on this site that explains in a little bit of detail. I don't think enough detail to fully understand sort of is it really protecting all the data here? And again, it clearly isn't because stuff does get through because they're told to escalate it and that sort of thing. But what it's supposed to do according to this page is that it will look for things like names, addresses, phone numbers, email addresses, and pull those out and replace them with like a blank person's name or something like that.

Joseph: 11:29

Now, there may be different versions of this model, but this is the one they pointed me to publicly. So it does that. It also looks for sort of strings that may be like a bank account number or that sort of thing. And sort of like hard PII, like hard personal identifiable information that's like a specific string or a specific set of characters or something that's clearly a name, it can remove that. But like, I don't think it can remove context and like a model is not going be able to do that.

Joseph: 12:01

And if it did, the prompt will be almost useless, you know what I mean? And then it wouldn't be possible to then train on it. The last thing is that apparently it does work locally, which I think is interesting. Like you could download this apparently onto a server, onto a machine and then use it locally to redact information. It's like a side issue to the story, but I might try to learn more about that because it's an interesting approach that I haven't seen when it comes to privacy, which is not to say this is the answer, because you still have people reading people's ChatGPT prompts, but hey, that's an interesting approach.

Emanuel: 12:34

Yeah, and we know from our previous reporting that AI is not good at this type of context. I thought how a lot of the jailbreaking methods work. So let's say I am one of these contractors that is reviewing ChatGPT chats. Can you say generally like, what am I doing? What is the process like?

Joseph: 12:57

Yeah. So you log into this dashboard, you select what they call a task, which is, you know, a list of jobs to select. These people, as we said, are contractors, they're like almost gig workers in a way, although I think that some may be paid per task and then others may be paid hourly. That's a little bit unclear to me. But they log in, they select a task and they see the prompt from the ChatGPT user.

Joseph: 13:29

First of all, the contractor has to summarize what do you think the user is trying to get out of this prompt? And an example given in the instruction guides I was given was that, oh, this person is trying to make an informative and good reply on Slack or something like that. Like, at this point, you are training, the contractor is training ChatGPT to recognize what the prompt is even asking for in the first place to improve that bit. So they do that, then they go on to the next bit of the task, which is that the dashboard provides, you know, three or four responses that ChatGPT has generated in response to this prompt. And the contractor then has to sort of review and rate them, whereas like, well, this one has a score of one because it is complete garbage and completely not usable, or this one has the top score of seven, which is like, is an excellent prompt and that sort of thing.

Joseph: 14:30

And they have all of this criteria for sort of rating these generated responses. And there's stuff like, if there are too many emojis, it should get a lower score. I'm sure that everybody has seen on LinkedIn or basically any social network at this point that you can tell if someone's using AI because there's a bunch of fucking tick emojis or whatever in their LinkedIn, know, Thinkfluencer post or whatever. They're trying to get rid of that. So they're downvoting responses like that.

Joseph: 15:01

ChatGPT doesn't want that. There's also, I think very interestingly, and I'll just ask Sam about this in a second, that the instruction guys specifically say, if the response is too sycophantic, it's not good. Vote it down. If it anthropomorphizes too much, like ChatGPT's responses saying, oh, when I was a chef in the French Riviera or some shit, like that needs to go down as well. And obviously, that is very, very interesting, especially the sycophancy stuff, because OpenAI has had a bunch of issues with real horrible tangible impact on people's lives because the chatbot was too sycophantic.

Joseph: 15:45

Sam, just briefly, because you wrote about this recently, what do you make of sort of them trying to improve the sycophancy and how does that relate to your story you wrote recently about someone taking their own life unfortunately?

Sam: 16:01

Yeah. I mean, just for a little bit of context, in April 2025, OpenAI removed a specific model called four o that was showing to be too sycophantic. And the reason that they said it was too sycophantic was people were harming themselves because they were getting too attached or, you know, we're we're believing everything that Chatbot said. And it was kind of it was it was propping up a lot of delusional thoughts is what they found. And overnight, I was like, woah, this is bad.

Sam: 16:33

I mean, the reason they said, woah, this is bad is because they started getting sued by the families of people who had died by suicide because of the the chat bot or, you know, allegedly. So, yeah, they they removed it in April, and then a couple of weeks later, they brought it back because people missed it so much. You know, it's like the the sycophancy and the reason that it still is a struggle, and it's interesting to see it continue to be a struggle in your story for them is it's why it works. It's why people use it. A lot of people use it.

Sam: 17:08

It's not why everyone uses it, obviously, but, like, it is kind of the the special sauce of ChatGPT is that it is this really conversational, warm, validating presence in a lot of ways. And, you know, they'll they'll argue that it's it's not meant to be, and it's not supposed to be and all this stuff, but that is what it is. And that's why people use it. And if they get rid of that feature of it completely, like that aspect of it, people wouldn't be so attached to it. I hesitate to use the word addicted, but like that is kind of the that's what we're getting close to in these cases is people getting really dependent on this chatbot that that gases them up and as a yes man.

Sam: 17:53

So, yeah, it's of course, they're trying to fix it because they can't just kill it. They can't just say, actually, we're gonna get rid of all of the features of that of the personality, so to speak, of ChatGPT and make it less sycophantic or make it not sycophantic at all and just make it like Google because people use it for that reason instead of just like googling stuff. They they'll actually use it because it does have that kind of very warm human back and forth.

Joseph: 18:26

Yeah. And so instead of the sycophancy, the instruction guides that I've seen, they want, as in OpenAI, want the chatbot to be more, quote, helpful, honest and truthful, empowering, and smart but humble. And then to close the loop on your question, Emmanuel, of what they do, after the contractors have sort of rated each response, they then have to like provide their rationale and their evaluation for those scores between one and seven. And sometimes that's a line, sometimes that could be a whole paragraph or something according to the guides that I've seen. But all of this, in case it wasn't obvious, is to improve ChatGPT's and OpenAI's models.

Joseph: 19:15

It is very much we are going to have humans read real user data, real ChatGPT prompts in order to improve the responses of ChatGPT. It's that.

Emanuel: 19:27

So what does OpenAI tell users about this happening behind the scenes? Does it tell them it's happening?

Joseph: 19:36

I mean, was a hard one because both you and I were searching around for like, their disclosure on this, right? And it's like, surely they tell people, they tell users that humans may review their prompts. So, you know, we found some examples where it will be something like, yes, we may review content to improve our models, we may hold on to data for thirty days and that sort of thing to review our models, etc. But we couldn't find at the time anything that was really explicit about there are humans who may read your stuff to improve the models. And I asked OpenAI this like straight up, like where do you disclose this?

Joseph: 20:26

They didn't respond. It was only after we published that they then sent a link to a page on their site, which I'd like to hear what you think, Emmanuel, but I think it was frankly buried that we couldn't find it, where they say, yes, humans may review data to improve the model. So like that's kind of the extent of what they get into. What what did you think of that disclosure?

Emanuel: 20:53

It's so buried OpenAI couldn't find it. How buried it is. Right? Like, we asked them, and they were like, they sent us some things that didn't explicitly say that. And then after we published the story, I think they were like, oh, woah, woah, we found we found like, it's like the fourth clause, sub clause of a different bullet point.

Joseph: 21:15

Yeah. Because it's not in the privacy policy, and it's not in this other thing. It's like in some help page. FAQ, yeah. Yeah.

Joseph: 21:23

So, you know, obviously we've updated the article with that so people can see that like technically is disclosed at some point. But like it was I don't think it's reasonable to think. Well, first of all, I don't think it's reasonable to think that most of ChatGPT's 900,000,000 users expect this to be happening. Yes, we're technology journalists. We know that humans are reviewing all sorts of stuff.

Joseph: 21:49

Probably our listeners know that as well. If you're listening to a very tech focused in-depth podcast, like you probably know that as well. And I think the vast majority of ChatGPT's users know that. And I think it's then almost like doubly unreasonable to expect them to find this really buried disclosure.

Emanuel: 22:08

You know who is not confused about somebody reading their chats behind an AI agent, a chatbot, is people who use Google Gemini, right?

Joseph: 22:19

Right. Because they explicitly say, and this was a I think I saw this on our own Google Workspace annoyingly. I keep trying to turn off the AI features and it comes up with like, every time I try and turn off, it comes back. But I saw some sort of disclaimer either there or on Reddit or something where Google says humans may review some chat content. And like it's pretty prominent.

Emanuel: 22:46

I think it says that sometimes in the box where you type the prompt, so it's unmissable. Right? That's like a more straightforward way to disclose it. Speaking of disclosures, we did find disclosures in OpenAI's documentation that says they review chats for safety issues, but this is different.

Joseph: 23:09

Yeah. Yeah. Yeah. So the safety issue is one that's been known for a while. That's like being part of the news cycle a few times now, of course, when unfortunately a series of people died by suicide, as Sam said, with the four-zero model, various measures came into force where ChatGPT will proactively monitor chats for, you know, dangerous behavior, problematic behavior, something where somebody's safety could be at risk, you know, and that could be escalated proactively, then a human may review it and then maybe they'll take action there.

Joseph: 23:46

There's other cases as well where somebody, you know, may be trying to plan a mass shooting or something and that they're conversing with ChatGPT to do that. And, you know, that will get proactively monitored, then raised to a human and then escalator and that sort of thing. So like human review has been part of ChatGPT's sort of pipeline and workflow for a while and in like some pretty crucial contexts, you know, where like, I think you probably want a platform as popular as this to have some sort of safety monitoring, right? So people aren't going to do a bunch of really horrible dangerous things and plan it on the platform, that sort of thing. And there's also, there was a Time investigation a while ago that was really good that found, you know, Kenyan workers working for $2 were doing trade or were helping with data annotation of, hey, here's a bunch of text, label this, and then we'll try to make the platform less toxic, that sort of thing.

Joseph: 24:52

All of those things are interesting and have been reported. This is different. The stuff we're reporting is different because it is firstly real ChatGPT user prompts. As far as I know, the Kenyan stuff was like third party texts that was being annotated. This is real ChatGPT prompts.

Joseph: 25:11

And this isn't about safety. This isn't about stopping a mass shooting or alerting the FBI if somebody is about to do a really serious crime, they're planning on ChatGPT. This is just to make the models better and to make the responses from ChatGPT better in OpenAI's eyes. So, know, presumably more people will use the product. It is different and I found a lot of people getting sort of confused by that.

Joseph: 25:35

I obviously think it's quite clear, but like this is a different thing that's going on and that's one of the reasons we're, you know, reporting it.

Emanuel: 25:42

Can we opt out of this?

Joseph: 25:45

That's the really interesting thing to me is that beyond the sort of disclosure that this may be happening being really buried on OpenAI's website is also how it is handled opt in or opt out wise. So if you go into your ChatGPT settings, if you use ChatGPT, obviously the best way to avoid this is to not use an LLM. Because I should say Anthropic confirmed to us that they're basically doing the same thing for Claude as well. It's just we happen to get the OpenAI related documents and that's why the piece is focused on them, obviously. But if you go into your ChatGPT settings, there is this one called improve the model for everyone.

Joseph: 26:28

And a little disclaimer on that says, allow your content to be used to train our models, which makes ChatGPT better for you and everyone who uses it. We take steps to protect your privacy. That's a very normal sort of disclaimer, right? Where it's like, oh, hey, do you want to share analytical data with us? Like you see that when you boot like any game on PlayStation or Steam nowadays, right?

Joseph: 26:52

It's like, hey, share telemetry with Versum will improve it or whatever, very similar to that sort of thing. Two things, this setting does not say that humans may be reading your ChatGPT prompts, like it doesn't say that. And secondly of all, this is on by default for ChatGPT 900,000,000 users. This applies to, I think the paid plan, the pro plan, the free one as well. The setting is off by default for stuff like the education plan, the enterprise plan, where you have companies who are rolling out AI across their entire organizations.

Joseph: 27:38

They probably don't want their employees uploading a bunch of trade secrets potentially or work related stuff, and then, you know, OpenAI training on that as well. So it's off by default for them. But for the vast, vast, vast majority of OpenAI users, is on by default. As far as I know, they're not asked to select it when they create an account, it is just there. And I know that's like really wild to me.

Joseph: 28:07

That's like one of the biggest mass automatic opt in mechanisms like I think I've ever heard of. Like, I'm sure there are bigger ones, but like this is like a huge part of it to me.

Emanuel: 28:19

Alright. I guess we'll leave that there.

Joseph: 28:22

Yeah. We'll leave that there. And then after the break, we're gonna talk about a bunch of a bunch of automatic stuff that Sam has been following. We'll be right back after this. Alright.

Joseph: 28:40

And we are back. Sam, here are a couple of stories that you wrote. The headline of this first one is Automatic CEO Matt Mullenwig put on leave of absence. I'm sorry. Did I pronounce everything correctly there, Sam?

Sam: 28:57

Yes.

Joseph: 28:58

Okay. Thank God. We we haven't had it in a while where, hey, here's the mispronunciation of the week.

Sam: 29:05

Sound it out. Sound it out.

Joseph: 29:07

We're All in on for years now, and I think we figured that I've I've kind of figured out how to say words nearly. We'll get to leave of absence, but could you please just remind people for those who don't know, like who is Mullenwig and like what is sort of a lot of the stuff that's been happening about them recently that sort of leads up to this?

Sam: 29:27

Yeah. Where to begin? Where to begin? So to try to do, like, a really quick, I guess, explainer slash too messy, didn't read. This kind of started and this being like the general crash out of Matt Milnewig.

Sam: 29:45

It started in 2024, and he started making these like big public accusations against WP Engine saying

Joseph: 29:56

that I'm gonna ask you to go even more basic. Like, literally, this is gonna sound so funny to you. What is automatic?

Sam: 30:06

Okay. Okay.

Joseph: 30:07

And what does it own? And what's its relation to WordPress? And then we'll go from there.

Sam: 30:12

Yeah. So automatic, no, that's good because I feel like people don't know that

Joseph: 30:17

I mean, I

Sam: 30:17

get

Joseph: 30:18

confused And their I read the articles, you know.

Sam: 30:20

I confused. I get confused. People get mad at me for being confused all the time. So it's been many years of trying not to be confused. Anyway, Matt founded Automatic, which owns wordpress.com, which is the commercial product of WordPress.

Sam: 30:38

And then there's wordpress.org, which is the open source project. So automatic does not own wordpress.org. Wordpress.com is the commercial entity that is owned by Automatic. Automatic also owns Tumblr, Podcast, couple other couple other things. But Matt is kind of at the at the helm of a lot of this in general.

Sam: 31:03

He is kind of the the driving force behind a lot of these these different companies. And what happens to automatic, what happens to wordpress.com, obviously, does have some effects on the greater WordPress open source community.

Joseph: 31:21

That a ton of people use because it's such a popular CMS, you know, for running websites all over the place. Yeah.

Sam: 31:29

Right. Yeah. It's they make they make lots of things that lots of people use for development.

Joseph: 31:35

Now how does it get messy?

Sam: 31:37

Yeah. I mean, it's already messy just starting from there. So yeah. So like I said, in 2024, he started making these, like, big public accusations of WP Engine. WP Engine is an independent company and one of those sites that hosts that host sites built on WordPress.

Joseph: 31:56

Like a hosting provider. Yeah. So, like, if you want a a website built on WordPress and you kinda don't wanna do all of the work yourself, you might go to this third party company called WP Engine, and they'll basically do all the work for you.

Sam: 32:10

Right. Yeah. So Matt was mad that WP Engine was he said that WP Engine was not credit like, was not contributing enough to the open source project part of WordPress and that they were using WP in the branding. So he was making the case that there's a trademark issue here where, WP Engine is infringing on WordPress's trademark. He in 2024, so this is two years ago now, he had a big talk during the big word camp conference, and that was kinda when people started being like, oh, shit.

Sam: 32:45

Matt is not okay about this. So that was when things started really getting, like, publicly messy is his feud with WP Engine. It gets weirder. It gets weirder.

Joseph: 33:03

Yes. Go on.

Sam: 33:06

So then in a little bit after that, because this all of this this kind of like public beefing with WPNJ and with anyone who didn't agree with him and, like, side with him completely on this issue. And he was like out there calling like WP antigen a cancer, a cancer to WordPress saying that accusing it that was like strip mining the ecosystem. You know, he was just saying a lot of a lot of stuff publicly and making a legal issue very in the public eye. So this created a lot of, like, turmoil within the company, within automatic, and also within the the WordPress open source project. So he decided to make an alignment offer to his employees and said, you agree with me and you're going to be with me in this legal mess or get out.

Sam: 34:01

And it was a buyout offer. And one and fifty nine people took that offer. And then he said, oh, I feel so much later. He wrote this in his blog, and then I think there was another buyout offer at some point. And then after that, at some point, he was like, oh, now we don't have enough people.

Sam: 34:17

Oh. So was like, oops. And the result of this moment in 2024 was basically what you have left automatic at that point was either people who were like completely, you know, like siding with Matt no matter what and kind of yes manning him on or people who were scared to leave or couldn't leave because of financial reasons and were staying silent. So that's what you have left automatic at that point. People were feeling very scared, very confused about their jobs.

Sam: 34:52

He kept doing weird stuff. Like, he he made it so to log in to WordPress, you had to agree that pineapple on pizza is good. You had to make a check mark decision because this was part of, like, part of, like, the legal battle that was going on. A judge ordered him to remove a checkbox that he had added that required users to pledge that they were not affiliated with WP Engine before logging in. And, again, this is this is what's this was affecting the wordpress.org login, so not even the.com.

Sam: 35:22

So he's just, like, left and right trying to, like, create this environment where it's like everyone agrees with me or you're either with me or you're against me. You get out. Yeah.

Joseph: 35:33

Funnily enough kind of brings us to this current news cycle or this recent one where, again, I'll just repeat the headline that, you know, Automatic CEO Matt Mullenwink is put on quote, leave of absence. So I think and when when do we publish this? This

Sam: 35:51

was That was Friday.

Joseph: 35:52

Friday. Dude, I don't know. September 9. So you start hearing about this. What do you hear and, like, what specifically do you figure out?

Sam: 36:05

Yeah. So actually, I guess it was, let's say, because I was on a flight to San Francisco when this started going down, which is pretty funny timing wise.

Joseph: 36:14

Worst possible timing.

Sam: 36:15

To be headed to the belly of the beast. So I guess it was, yeah, it was Wednesday when I started getting messages from people saying people sources who are close to automatic sources that are obviously invested in this saying, Matt has posted in Slack in the announcements channel, which is the company wide channel for automatic, that he has been forced out. The wording specifically, and it's worth reading is because it was wild, was that the CFO, Mark Davies, and the rest of the board had conspired behind his back and voted to put him on a leave of a paid leave of absence. And he said he voted against that, And they didn't give him they gave him less than an hour to have it reviewed by legal counsel. And eventually they voted to it a it was a coup is what you're saying.

Joseph: 37:15

So this I mean, it sounds like some succession shit. Yeah. Basically. I mean For sure. Not not as not as cool as succession.

Sam: 37:22

No. But you definitely hear the you hear the you hear the piano. It's it's starting. Think you

Joseph: 37:27

can say it's or something in this case. I don't think it's that very cool.

Sam: 37:32

Yeah. So succession thing, but on a banjo.

Joseph: 37:34

Yeah, yeah. So that happens and you get it out and you know, you break that news and then a bunch of other people start following it and doing their own reporting. And it seems that, oh, okay, well, that's the end of the story. No, like forty eight hours later, something like that, we published this other one. The headline is automatics Matt Mullenwig claims he's back quote, in control, end quote.

Joseph: 38:04

So what changed?

Sam: 38:07

So, yeah, so this is what happened on Friday because I started I was on, at that point, West Coast time and was getting screenshots and messages at, like, 4AM. So I just remember being in this hotel room, like, in the dark, like, was trying getting blasted by my phone screen. People saying that Matt's back. So Matt, again, obviously, he's doing he's airing this out once again so publicly, which is we can get into, like, whether or not that's a good leadership style. I'm sure he thinks he's being transparent or something, but it's stressing people out.

Sam: 38:39

And people were already, like, stressed out because he just was like, I've been ousted. And whether you are happy about that or not as someone who works there, it's destabilizing. So and also the board had not really said much publicly. They were kinda like, yeah, we're automatic sent a statement to us and was like and to others saying, you know, we we acknowledge that Matt is taking a paid leave and that Mark Davies is gonna be the interim CEO. So on Friday, Matt posted in announcements again, happy to announce the board is back in agreement, and I'm in control of automatic.

Sam: 39:17

A lot has happened in forty eight hours that we need to sort out, and I hope much of that was a misunderstanding because I have huge respect and regard for those involved. So yeah, people were like, what the fuck at this point?

Joseph: 39:29

Well, and you had heard about it before when we post that blog post.

Sam: 39:35

Yeah. Yeah. Well, so there was a blog post which was not even related. So what happened was he posted that in Slack publicly. Everybody saw that.

Sam: 39:43

He had been posting unhinged shit on Twitter in the meantime, talking about, like, coup attempts and things like that. And then on Friday morning, he told me and TechCrunch that he was going to publish a blog soon, and he's never replied to me from his personal email ever until Friday. And I've sent him a lot of emails. So I was like, okay, this must be something. So the title of the blog post, which is on his personal blog, ma.tt major life announcement.

Sam: 40:20

So it's like, he's gonna talk about the mass, the coup. And what the blog is about is about him buying a tugboat. And it's all about buying a tugboat. It's just about buying a tugboat. And you know, he replied to me, and he was like, the blog is up, and it was linked to that.

Sam: 40:39

I was like, you are such a troll. It's such a troll that I get that I always have to respect it. So I would respect it if it wasn't stressing everybody out so bad. So that was kind of where we left it on Friday. I obviously emailed everybody again and was like, can you please give us a statement or something?

Sam: 40:59

Automatic did send a statement, which I can just read because I don't wanna get it wrong. But that was I was trying to confirm with them that he was back because, obviously, we've heard a lot of different things. So on Saturday, they said, Matt Mullig is the chair and CEO of Automatic with full support of the board. And if you search online, you can see many top executives and automaticians supporting him as well. So the statement was basically like, you know, he's back.

Sam: 41:29

Everybody loves it. Yes. Which everybody does not love it. No.

Joseph: 41:36

No. Not everybody loves it. But then also, I think a day or so after TechCrunch did this other story, which is sort of where we are now, I think, and their headline was sources say Autopatics Board is out after failed attempts to oust CEO Matt Mullenwink. So he's out on a leave of absence, somehow comes back apparently with the full support of the board that tried to oust him.

Sam: 42:06

That's the thing is like, that's probably why he has full support of the board is because it's not the same board anymore. Right. Right. Right. Right.

Sam: 42:14

And now it's like It's different board.

Joseph: 42:16

Yes. These people have gone. So where are we now just based on the last TechCrunch report and sort of your reporting? Like, so what's real? What's going on?

Sam: 42:27

I mean, at this point, it's like it's all still very story is still developing as they say. We know that Mark Davies is out.

Joseph: 42:38

He was the CFO.

Sam: 42:39

Right? He was gonna he was the CFO. He was gonna be the interim CEO for for two beautiful days. He was interim CEO for a beautiful room, I'm sure. Sure.

Sam: 42:50

Probably being truly stressful for everyone involved. So he's out. TechCrunch reported that they had gone through and looked at like the social media accounts of a bunch of different board members. And you can go read that because it's a it's a good story, but it's also like it's more complicated than when we wanna get into on the pod. We found out that Tony Schneider, who is also he's the lead at Blue Sky resigned from the board.

Sam: 43:24

He wrote something about his his time at automatic on his own personal website. Sue Decker and Ann Dunwoody both either were removed or resigned according to TechCrunch. So the old board is gone. New board. TBD.

Sam: 43:45

Actually, let me look and see if there's anything about the new board on up here live on the on the pod.

Joseph: 43:52

Well

Sam: 43:53

I doubt it. Yeah. It's all the same. It's all the same people still on that site, on automatic site. So they haven't even updated it.

Sam: 44:00

I don't know. It's just like it's constant chaos with Matt these days. So we're following it obviously, but I think things are still very much up in the air. The dust is not settled. So Yeah.

Sam: 44:19

That's where we're at.

Joseph: 44:20

Makes sense. Alright. We'll leave that there. If you are listening to the free version of the podcast, I'll now play us out. But if you are a paying four zero four Media subscriber, we're gonna do what we do every week, which is talk about another one of our stories.

Joseph: 44:35

You can subscribe and gain access to that content at 404media.co. We'll be right back after this. Very stupid little car in the background like, Someone one of you's got a fucking clown car out there. Alissa, keep keep this in. Alright.

Joseph: 45:02

I I like that as a running bit. I'm sure every podcast does that where they're just like, Alissa, keep this in. Well, they don't all have that name. Anyway, so we're back in the subscribers only section. Jason was gonna join because he's having some Wi Fi issues.

Joseph: 45:18

So we're going to do one of Emmanuel's stories. The headline of this one is a 16 z says you actually love social media and inshitification isn't real. I mean, where do you want to start, Manuel? I feel that everyone knows what in stratification is, but like, do you want to define that or what do you think?

Emanuel: 45:40

I think everyone knows, but for the record, it is a term coined by Cory Doctorow, and I believe it is one of the stickiest and most explanatory concepts of the past five, ten years about what is happening to our lives because of mostly technology, but I think like every private company and the theory is that a great example for this is Uber or Netflix or Amazon Prime, where there is a period of time where these companies are trying to amass a huge audience. They subsidize the services. The services are very good. Netflix is very cheap. Amazon Prime is very cheap.

Emanuel: 46:34

It works very well. You get your packages immediately. Uber is much cheaper than a taxi, and then once they have huge portion of the market, something nearing monopoly, they have a captured customer base, and then they begin to squeeze it more and more and more. The services get more expensive and they get worse, and that is what instantification is, and that's what we're all experiencing in every aspect of our our life as consumers. That's basically the concept.

Joseph: 47:12

And I feel it's almost like obviously true. If you just like look at your screen or even out of your window, you know?

Emanuel: 47:20

It is true in the sense that you can quantify it in various ways, like you can track the prices of Netflix, you can track the prices of Uber, you can count the many ways in which these services got worse, but I believe the reason that it is such a sticky concept is that you feel it deep in your bones. Like we live in such a consumerist society. We use these services every day. They're such a big part of our life, and we all feel everything getting shittier around us, even though these companies are making more money than ever, and they should have the ability to provide better services, and we all just are walking around with like this deep, deep rage about, you know, gentrification. Just as a personal note, like for me, the like clothes and furniture are like some of the most gentrified product types there are.

Emanuel: 48:22

Like, if you're looking for like a good t shirt, there's like maybe two companies that make t shirts that will last you a year, or if like you're looking for a couch, you have to spend, I don't know, like more than $6,000 to get like a good couch, and that that's like a relatively relatively new thing.

Joseph: 48:42

Yeah. You don't fucking go to Wayfair. I fucking hate that stuff. And it just falls apart instantly. You're like, well, now I have to buy another one, and now it's like actually more than double the price.

Joseph: 48:52

Anyway, whatever. Okay. I'm gonna I'm gonna get

Emanuel: 48:54

See, you're mad.

Joseph: 48:55

I'm gonna get legit mad about it. So, we feel it in our bones, but a sixteen z published an article arguing that you are wrong and you are stupid for feeling that and you should shut up. That's not an accurate paraphrase. What were they who So wrote this and what's their beef in stratification?

Emanuel: 49:22

So H16z is the venture capital firm that is a huge player in Silicon Valley. It has powered Facebook and and Airbnb and a ton of these companies, still a major player, a major a major player, NII, and so on. Extremely powerful more than ever because of its links to the Trump administration now. And they have they're they're kind of for the past three years, five years are in an all out war with media that's critical of technology. And one of the ways they have done that is they kind of have their own media arm, and that is where they published this article, which was written by New York University professor.

Emanuel: 50:20

I'm sorry if I'm mispronouncing his name, but it's Ruby Telo, Telo, something like that. And basically he argues that in gentrification, this very sticky concept we just talked about, is not real. Like he says, or the argument he makes is that the evidence shows that this has not actually happened, people don't actually feel this way. What is actually happening is that these services are getting better and what he he coins this new term, which I think is interesting. Like, is it is a thought provoking argument.

Emanuel: 51:05

He calls it platform narcissism. And essentially, he uses social media as the test case. And everybody talks about how much they hate social media, but more people are using social media than ever, like the numbers just keep going up. He points to some data that he claims shows that people actually are enjoying these services more than they did in the past. We'll get into that in a minute.

Emanuel: 51:37

But he says that the real reason for why people say that social media is bad is not that everyone feels this way. He says there's a particularly loud, minority of people in the world, and these are elder millennials, of which I am one. Not to blow up his spot, I believe that Corey might be like a very young Gen Xer. He doesn't quite qualify as an elder millennial, but this article claims that he is. And what is happening to them is they grew up with this technology when they were young.

Emanuel: 52:18

This technology has served their needs. They grew accustomed to like a certain version of it, and now they have aged out of it. The technology is serving a completely different majority of users who are global, who are younger, who have different needs and wants from social media, they are actually happy with it, and everyone who says that social media has been shitified are just like old, crotchety, out of touch people who should like quote, be focused on raising their kids, rather than being online and complaining about how social media sucks now.

Joseph: 52:54

Right. And are the specifics about this like, oh, you aged out of Facebook and now people love TikTok, or is it you aged out of Facebook and actually people still love Facebook? Like, do they give specifics like in the argument or is it?

Emanuel: 53:11

Believe that there's like the example that he gives is that we we being millennials used Facebook and use a lot of like text and commenting and threads and stuff like that. And what Meta has done is essentially shift towards Instagram and Instagram Reels particularly, which is like a more visual algorithmic video focused service that people actually love, which is supported by the fact that so many people use it. And again, he points to this data that he says says that people actually like it, which I'll get to, but like, what do you like, what do you make of the like platform narcissism argument? Does that track for you at all or?

Joseph: 54:05

I think it's interesting as well. Like I'm not saying I obviously agree with the rest of it. Like I still think it's justifications real. I think that's like a separate thing. The idea of platform narcissism and that, you know, people can almost be like clouded by their own experience of maybe social media and that, you know, maybe they should have they've aged out of it.

Joseph: 54:27

That's what I think there's probably something there. I just don't know if it actually carries the argumentative weight that the author thinks it does. But I think it's also probably still a thing. You see what I mean?

Emanuel: 54:42

Yeah. Personally, like, I never clicked with Facebook. Like, I was never a Facebook user. I don't know about you guys, but it's

Joseph: 54:53

Oh, yeah. Used to I mean, I used I mean, MySpace is going too far, I think, because that died before we can get shitified, but like it's and that wasn't really a thing. But yeah, I used Facebook for sure. And then, I don't know, even like, like other platforms as well. Yeah, absolutely.

Joseph: 55:12

But like if you didn't click with Facebook in the first place, it kind of almost doesn't apply to you in that instance.

Emanuel: 55:17

Right. Like the only social media I actually used was Twitter, and I stopped using it before Elon Musk took it over, and now I have, I use quote unquote blue sky, which means I read it, I barely post, and I guess, I don't know, I am following the article's instruction, and I am focused on raising my kids and not on my posting. So I guess I do fit that mold anyway. But I definitely, I don't know. It's like, I I left Twitter.

Emanuel: 55:49

Like, I look at Twitter a lot less because and I don't want to, like, objectively prove the argument, but it isn't shitified. Like, it's a piece of shit. It's a trash heap. I do not enjoy looking at it. It's a very upsetting, unpleasant experience to scroll Twitter.

Emanuel: 56:06

So let's let's get to the like, the evidence for like, people actually like this. This is really what the argument hinges on and is a very silly sleight of hand. So he points to the American Customer Satisfaction Index, the ACSI. It is like this big research firm. They publish an annual report that surveys a ton of customers across all service types and industries and so on.

Emanuel: 56:45

And they also survey social media applications, and that fits into the category of the ACSI entertainment study. So the one for 2026, which looks back at 2025, it surveyed a little bit more than 30,000 customers, which is a respectable sample size, like that is a good number of people to sample if you're looking for their sentiment about a particular service. And these people are asked a bunch of questions. They rate the services on a scale of one to 100 around various categories, and then all of that is consolidated and given an average, and that's how they score, what people think about any given social media. And basically what he shows, looking back at this survey, going back a few years, is that it keeps ticking up, and in the past year it ticked up 1%, right?

Emanuel: 57:56

So objectively, more people than ever are using social media, and he argues objectively, they like it 1% more than they did last year, so they don't actually hate it. There is no actual intrudification.

Joseph: 58:13

Overall satisfaction is the term in the piece, yeah.

Emanuel: 58:17

Yes. And I read that and I was like, well, that definitely goes against like every fiber of my being and what I feel and see that other people are saying about social media, how can this be? So I drilled down to what this ACSI survey is actually asking customers about, right? Like, how is it, what are they asking them to rate between zero and one hundred? And this is where the disconnect is, and that's because they're asking questions like, here's some questions that they ask the people who are participating in the survey.

Emanuel: 59:04

They're asking them to rate things like quality of mobile app, reliability of mobile app, including minimal downtime, crashes and lags, loading speed and reliability, ease of navigation, ease of using the site on different devices, speed and reliability of video clips, variety of services and information, freshness of content, relevance of content, ability to use the site without disruption.

Joseph: 59:36

It's like That's all tech shit.

Emanuel: 59:38

That's all, yeah. You're like a consumer surveyresearch firm, makes a ton of sense to ask these type of questions, right? Like if you are a reviewer for Wirecutter, right, and you're reviewing TVs, then it's like, what is the quality of the image? Like, how long does it last? Is there burn in?

Emanuel: 01:00:05

Stuff like that. It's like, does my toaster toast the toast? And it's like, it does. And social media definitely works. I can assure you that Instagram Reels loads very fast and good.

Emanuel: 01:00:19

You know what I mean? Like, it delivers videos to your face at a rapid pace. They're high resolution. The app loads. It almost never crashes.

Emanuel: 01:00:28

Like, all of this is true. On the other hand, you scroll reels for an hour, you're going to feel terrible. Like, it's not it makes people feel bad, right? And it's like, it is possible to ask people these questions and get their opinions about, well, how does this make you feel? Or do you want your kids to be using this app?

Emanuel: 01:00:51

Or do you think that this app has an overall positive or negative effect on the politics of your country? And people do those surveys, and I list a bunch of them in the article, and when you ask those questions, people hate social media. They think it's bad for the country, they think it's bad for their kids, They think it's bad for their mental health. It's just like a universally disliked, I would say mildly type of service along those lines. That doesn't mean the apps don't work.

Emanuel: 01:01:32

They work very well, which I would argue is part of the problem.

Joseph: 01:01:36

Yeah. Yeah. Absolutely. Sam, do you think intratification is real?

Sam: 01:01:44

I mean, yeah, it's obviously real. Did you know there are apps that you can download that are Instagram without the reels?

Emanuel: 01:01:51

I did not know that.

Sam: 01:01:53

Yeah. I mean, I think that's probably a sign that people don't particularly feel positive towards your biggest feature on your app.

Joseph: 01:02:04

So it's just a third party client, but it, like, blocks loading of the reels part or the reels page or

Sam: 01:02:10

It's like a light version of Instagram. I haven't tried these, but I just made I was just became aware of them, like, last week. Because as we know, I have a a serious reels addiction. Yeah.

Emanuel: 01:02:20

The fact that there's a whole cottage industry of services that are designed to prevent you from using those other services Yeah. I feel like does not speak well of them. There isn't like an application or service that prevents me from like using my toaster, you know? It's just like not a thing.

Sam: 01:02:38

There's not like only reels. I don't think. It's not there's not like more reels faster. You know? It's like intensify the reels.

Sam: 01:02:48

It's it's like we people feel like they can't put down the reels, and so they need Instagram without it. I mean, yeah. I also I use Facebook a lot. I was definitely right in smack in the middle of the millennials that would go out on a Thursday night, get wasted. And then Friday morning, the first thing you do is upload 40 pictures to an album called Thursday night of everything you did.

Sam: 01:03:14

And it was bad.

Joseph: 01:03:15

So painfully accurate.

Sam: 01:03:17

Or it would be something it would be like a like a lyric. So, yeah, I don't know. It's like, I I definitely was like a heavy Facebook user. I I will say Myspace, I was a heavy Myspace user.

Joseph: 01:03:29

Mhmm.

Sam: 01:03:30

And Myspace isn't shitified. It still exists. It's just shitified. It's shitty.

Joseph: 01:03:35

Well, didn't it become a music site?

Sam: 01:03:37

Yeah. And like it's it's impossible. It's like nobody uses it because it sucks. Yeah. You know, and Tom drifted away into the ether.

Sam: 01:03:44

Good for him. So, yeah, I I told Emmanuel when he when he was writing this headline, cause the headline is like, everyone hates social media or something. And I was like, I love social media. Actually, it's not everyone. I really like social media.

Sam: 01:03:57

I don't like what they've done with it.

Emanuel: 01:03:59

Right.

Sam: 01:04:00

I think that it's it which is the point of the article is, you know, it's what they've done with the place sucks. So yeah. Yeah. I don't know. Yeah.

Emanuel: 01:04:11

Do do you have, like, a most reviled, like, and shitified service or product type in your life? Like I said, for me, it's like clothes and furniture. It's just like, I feel like you can't buy anything good without spending a fortune.

Sam: 01:04:26

I think that's I don't know if this yeah. I agree with that for sure. I was in J. The other day, and I was trying to find just like a little like a linen button down shirt for less than $1,000,000,000, and it's all polyester now because they got they got like pee bought or something, but it's shitty.

Emanuel: 01:04:41

It used to be like, I'm going to J. Crew. That's like the fancy Yeah.

Sam: 01:04:44

And now was like, I'm gonna get like a real like, I'm gonna like splurge with a real linen shirt, and I literally couldn't. They don't they were like, we don't do that without some kind of like fucking blend, and also they're gonna be like $90. It was on sale for like $90. I was like, this is crazy. I don't know if this counts, but fuck.

Sam: 01:05:04

I forgot what I was gonna say. My brain is so my brain is my worst in should have find products. My own mind.

Joseph: 01:05:10

Your frontal lobe isn't should have found. Oh

Sam: 01:05:13

my god. Yeah. I'm just I'm just like the loading screen right now. Yeah. I would say that's definitely oh, I know what I was gonna say.

Sam: 01:05:20

The fact that I can't I rage about this so much. The fact that I can't go to the store and buy something really simple so much of the time. It's like, I can't it's hard to just go to the store and buy something now, not just because it's all locked up, but like stores just don't like clothing stores, especially won't have like all the sizes there, like go online. It's like you go into like a specific, like even like a department store and it'll be like, oh, we only have like these particular items that are very common. And if you want something different, it's gonna be on the website.

Sam: 01:05:53

Like, I was able to buy this 10 ago.

Emanuel: 01:05:55

Yeah. I found out about this actually. Sorry. Now we're on a tangent, but I found out I think it was sorry, I'm doxing, but we we went to buy my wife and I went to buy clothes for your wedding. Mhmm.

Emanuel: 01:06:08

And We went to the we went to the mall. We went to King of Prussia Mall, which is like one of the biggest malls in America or something, like very, very fancy. And we went to like every store, and they had nothing. Like, you go to like fancy brand name stores and like, they don't have clothes. And I found out that the reason for that is that the profit margins are all in accessories.

Emanuel: 01:06:33

They're like in sunglasses, purses, and stuff like that. So that's what they have in the store because it because it's like, it's literally like the profit margin per square foot is all in accessories. Yeah. And if you want anything else, they're like, yeah, do it online.

Joseph: 01:06:48

We'll

Sam: 01:06:48

Yeah. Ship it And it's like, with clothes, especially, I'm like, I hate like, shopping as a woman is hard, but it's just like, I don't know, like, brand to brand, I don't know what's gonna fit right. And then I have to buy it online because I can't go to the store and just be like, I wanna try this stuff on. I have to buy online to try a bunch of shit on. And then I'm just like, oh, I'm like, I'm destroying the planet actively because I'm getting stuff mailed to me, and then I'm mailing it back, returning a bunch of shit that I couldn't try on the store.

Sam: 01:07:18

It's just like, I don't know if anyone has a solution to this other than just like live extremely close to like every store. I don't know. But it's like, I live in New York City. There are stores and I can't just go in and like buy something very I was trying to buy like pens the other day. And I was like, where the fuck do I buy a pen?

Sam: 01:07:34

I had to go to like a stationary store. It's like, this is and they had like the fancy pens. It's like, okay. I guess I'll do that.

Joseph: 01:07:40

The stationary store is gonna be a boutique fancy Japanese stationary store, and it's gonna cost you a bazillion to get a pen or whatever. And to be clear, those are good stores, but, like, you just wanna buy a pen.

Sam: 01:07:53

Yeah. It's like, I just where do I go? I don't know. I feel it feels very it's a silly problem, but it's also like like you said, it's like just buying something as simple as like an outfit for an occasion is so hard.

Joseph: 01:08:07

Yeah. Alright. I think this turned into a Seinfield stand up set.

Sam: 01:08:13

Yeah. This is

Joseph: 01:08:13

a What's millennials' beach with that? What's up with airport food? Okay. So that's good. We will leave that there and I'll play us out.

Joseph: 01:08:26

As a reminder, four zero four Media is journalist founded and supported by subscribers. If you wish to subscribe to four zero four Media and directly support our work, please go to 404media.co. You'll get unlimited access to our articles and an ad free version of this podcast. You'll also get to listen to the subscribers only section where we talk about a bonus story each week or this week of bonus moan series of complaints. This podcast is produced by Alyssa Midcalf.

Joseph: 01:08:54

Another way to support us is by leaving a five star rating and review for the podcast. That stuff really, really does help us out. This has been four zero four Media. We'll see you again next week.