What we're looking for is where can AI help you be more human rather than less human. So the first round of AI slop was all about being less human. It was fake ideas, fake concepts written for other AI algorithms. And it didn't reflect well on the creator and it didn't help the reader.
Hello and welcome to the Media Co-Pilot. It's a podcast that explores how AI is changing media, journalism, and content creation. I'm Pete Pell, longtime tech journalist and founder of the Media Copilot, and I'm excited to bring you fascinating conversations with fellow journalists, media executives, and the people building how the business of news will work in the future.
This week, I'm thrilled to talk to Tony Stubblebine, CEO of Medium. Tony took the helm of Medium in 2022, and since then, he's been guiding the platform through a pivotal transition, moving away from a publication model and refocusing instead on independent writers with the goal of creating a healthy, sustainable space for them and online publishing in general.
Now, Medium's taking that mission into the AI era. The company just joined a coalition of major publishers and tech companies that includes Reddit, Yahoo, Quora, The Daily Beast, and they're all supporting a new web standard called RSL, short for Really Simple Licensing. The standard's designed to give creators and publishers a way to clearly define how AI systems can use their content, and importantly, how they should be compensated when they do.
Tony and I talk about RSL, of course, but we quickly go way beyond that. Tony's a very deep thinker about writing and what its role is in how we get information online, and most importantly the incentives that shape how people build businesses around that. Be warned, we get a little philosophical here and there on what AI means for our media ecosystem. But it's all in service of figuring out what the rise of AI means for the open web and for writers everywhere.
So, I hope you enjoy the conversation and if you do, I'd be grateful if you'd leave a review for this podcast, especially in Apple Podcasts and Spotify. They really do help the show. Also, a couple quick notes: I'm offering a monthly class on AI for beginners. It's only 60 minutes, focused on practical takeaways for media professionals in journalism, PR, or communications. It's only $49, it's a live class so it's always up to date, and I guarantee you'll walk away with something you can apply to your day-to-day. Sign up at mediacopilot.ai.
Also, this is your last chance to enroll in one of my two advanced AI courses starting in November. One's for journalists and the other's for PR and comms. These are loaded with lessons on things like deep research, how to automate workflows, and even building your own agent. Sign up at mediacopilot.ai.
All righty. Now it is time for my conversation with Tony Stubblebine, CEO of Medium. Tony, welcome to the Media Co-Pilot.
Okay, Pete. Yeah, thanks for having me.
I'm thrilled you're here. We need someone who's in the thick of all this AI business, particularly how it's intersecting with big platforms and in particular creator-driven platforms like your own. I know you've recently talked about RSL and all of that, I want to get into all that. But right now I want to zero in on Medium itself a bit because obviously it's gone through a bit of an evolution over the years. It had publications for a while and it's always seemed to have evolved to memberships and now obviously you've taken some stances on AI and licensing. When you look at what Medium is today, how would you describe it and what is it trying to be as we're advancing to this next chapter where AI is a big factor?
Medium is a writing platform for everyone and I think that kind of got lost in a certain era of the internet where we all got obsessed with influencers and content creators and kind of quitting your day job to be an independent media creator. But what got me excited about the internet was the idea I could hear from anyone, and especially if that anyone happened to be an expert in some big or tiny topic that I cared about.
Medium in the many iterations of Medium, it went through a period where we did have a lot of professional media people on Medium, sort of like yourself. And we've gone back to being a place for amateurs because I think a lot of times a professional who writes is a very exciting person to hear from.
You take our business essentially—the part of the business that makes Medium profitable is people writing about their job so that people who want that job can learn how to do their job better. We have a huge traction in, I think what we're going to talk about right now actually, is a lot of people are coming to Medium to learn how to use AI to do their job better. That'd be an example of professional development, which is a core pillar inside of the Medium community.
Do you guys use the term UGC anymore at all? Is that kind of how you describe it?
Totally. I use it because I grew up in that era. I was working for a book publisher in 2003 when we realized everything we were writing about was a year too late and had already been covered by some UGC blogger on the internet. And those bloggers were getting into the crevices that we would never bother to write a book for. I think that term is so technical and so boring, but I haven't found a better one. I might test one on you if you want. I think of it actually as, if you're the fourth estate, Medium represents the third estate. Typically religion and politics and government, they're the powers, and the press holds them to account. And the third estate was the commoners. They never had a voice. And this is what the internet changed—the idea that all of us could have a voice.
It turns out sometimes that means you're hearing from the same person you would hear from at your local bar, but a lot of times it means you're hearing from some expert that was otherwise unfindable before. And so we try to narrow in on that as the place where we do the most good and we're the most helpful.
So to someone in the industry looking at the broad categories here, I might put Medium at least in the same bucket as Reddit where they're both UGC platforms. You could almost say Medium's the cleaned-up version of Reddit that's much easier to navigate. But Reddit is an interesting story in the AI era in that it somehow became newly valuable for its corpus of human-generated content. The AI platforms became valuable and they've leveraged that more effectively than a lot of people. Do you see a similar path with Medium, and has that led you to some of the things you've been doing lately, in particular RSL? How do you see yourselves vis-à-vis that platform specifically, and how has that affected your strategy with regard to AI?
We're going to talk about the RSL standard. I love it. You led with media needing saving and we're going to talk about an internet standard—famously popular and viral way to grab eyeballs and attention.
Yeah, I guess maybe we should talk about the RSL standard too and what the hell that is.
We could talk about the problem essentially. The AI companies launched in a way that was antisocial. Our society is based on an exchange of value and they launched without any exchange of value. They didn't ask if they could train and then they didn't offer any money and they didn't offer any credit a lot of times.
The internet runs on a particular type of credit. It's that I link to you and send traffic to you. We all let Google train on our content because Google sends us all this traffic, right? So that's the social contract that the internet was based on, and the AI companies ignored that. In a way I feel sorry for them because if they don't come back with a working social contract, the internet is just going to retreat from public publishing altogether.
Public publishing is weird, right? Because you're putting yourself at so much risk right now from getting attacked by strangers and misinterpreted. It can be very toxic. But the thing that makes it worthwhile is that Google had been offering so much free traffic to put yourself out there. But if you don't have that incentive, we're all just going to retreat to private spaces again.
You see this in so many trends right now. Substack is an example of a paid newsletter that's your semi-private space. A lot of the Twitter exodus went to Discord, definitely went to private group messages. Over the summer we spun up a whole separate blogging platform to experiment ourselves, based on the idea of private blogging, and immediately you realize how many things people want to say that they would never say to the entire world.
There's something the AI companies have missed: if they don't provide the incentive for people to publish in public, their ability to train and learn will just flat out disappear. And so the content world—people like Medium, Reddit, Quora, O'Reilly Media, which is actually a company I used to work for—have gotten together to support a new internet standard.
This always sounds so boring to me, even though I guess I was early and excited about it in some wonky way. It's called the Really Simple Licensing standard. It came from some of the people behind the RSS standard way back in the day that made blogging and podcasting possible. The idea is to have a standardized way to talk about this tension around the social contract.
If you as an AI company want access to the content we have, here is the arrangement that we will allow. Otherwise, we'll block it. Medium is capable of poisoning the results too. You've actually seen Cloudflare will poison the content that AI crawlers see as a way to dissuade them from being a bad actor.
Talking about their blocking by default?
They have something beyond that that will just rewrite some of the content.
So essentially, if they're bad actors, we have tools to deal with them in the same way we have tools to deal with spammers and whatnot. I think there's going to be a way to force the AI companies to some sort of agreement. And that agreement is either going to be you got to send traffic or you got to send money. And if you don't, why would we support you in any way? We're going to hide ourselves from your view.
Because there wasn't a standard, every media company went and did a custom deal. Reddit was very early and successful about doing a custom deal. Medium, I think we were just waiting for the standard and thought it was better for the internet if we all operate together.
As opposed to all these licensing deals which sort of was the way things have been done for the past couple of years.
Exactly. One of the problems is that if you do ad hoc deals, you don't have the negotiating leverage. What you mentioned about Cloudflare and their blocking service is the smart thing—they realized in order to start the negotiation, you had to be able to threaten to turn off access to almost all content. If one media company blocks their content, there's still plenty for the AI companies to train on.
So you needed leverage first. It doesn't make sense to me to have vendor lock-in—I don't want to have to be a customer of one company in particular in order to negotiate with all AI companies everywhere. So I really appreciate what Cloudflare did. We're very happy customers of Cloudflare, but this is the thing that deserves an internet standard rather than to be offered by one.
So if I'm a Medium writer or one of your users, how does this manifest to me? Is it just a setting that I go in, what is the mechanism here and what can I expect?
Our intention is to start the negotiations and then pass all of the money back directly to the authors. I think we're the only social media platform that thinks that way—that we're just negotiating as a service.
The first step of the negotiation is to threaten to take all the content away. And so we've been more aggressive about blocking and I think starting those negotiations, but there's nothing really for the writers to do until AI companies start standardizing on what they're going to pay.
I'm 90% certain within the next year there'll be some AI payments flowing through Medium to the writers on Medium.
How does this sort of fit into the payments ecosystem for AI, which is to say there's startups like Tolbit doing monetized endpoints. Maybe this is more for people on the open web as opposed to Medium. There's people who do essentially pay-per-crawl or pay-per-inference. We've been talking about Cloudflare doing it at the CDN level, and we're already throwing all these acronyms around. To be honest, I'm in this and I can't keep it straight. What if I were a little guy who has very little resource, whether I'm on Medium or not, to even think about this? I just want to turn something on and get paid whatever the fair rate is. Before that, where does this fit in and are all these systems compatible, or will something win out?
I think there's going to be two eras is how I would answer you here. You're talking about wanting the standardized era where everything is simple and there is a known rate for what people get paid. We're not in that era. We're in what I would call the 'get your popcorn' era because it's a bunch of big companies fighting and something is going to come out of that. There is no standard rate. What we have right now is the first part of the fight where we have a standard way to say no. And once we have a standard way to say no, we can open the conversation about how we're going to say yes.
This 'get your popcorn' era includes lawsuits. It includes poisoned AI. The em-dash is popular in AI-generated writing in a large part because the em-dash is very popular on Medium. We were one of the big training sets, and it was the first time we realized we're big enough to change the output of an LLM.
Let me just double-click on that really quickly. Given that, and assuming you're correct—to my knowledge, Medium has not entered into litigation on any of these LLMs. Is that correct? Why not, given everything you've just been saying about compensation and how that should broadly be handled?
You're getting into essentially the explanation I give people at work whenever we're talking about legalities. Anytime you're in a legal situation, you also have other levers—one is social and one is power. I look at this situation and I say I don't think the legal standing's actually that good. It's not turning out necessarily to be a copyright violation. You might have some standing around terms of service and crawling, but the original training—I'm not sure that it was illegal. And that's a long battle for us.
Importantly, where's our standing? We don't own the copyright for anything on Medium. All of our users own that copyright. So you're talking about us as a software company—do we get in the business of facilitating lawsuits for our customers? It's very far away from what our actual business is.
It's an important distinction. I get it.
Right. But what we do have is power and influence. And so that's what we try to wield. The power is essentially what do we block, what do we display. And the influence is how can we help stand up internet standards, because I think the small AI companies will ignore the standards but the big ones will find that they really have to participate.
Let's talk about standards. You mentioned in about a year you would expect something to coalesce either around a standard or just a new way of doing things from a business perspective. Do you think the courts are going to be a part of that in terms of steering things, or can that happen absent some indicator from courts or regulatory bodies? And I wonder about vis-à-vis that—as more sites like yours do blocking, as players like Cloudflare enter into this, and you mentioned the poisoning, does the quality of AI outputs start to go down if they don't come to the table? Is that going to be the thing that forces them?
