PodQueue

There are huge gaps between the human needs of digital-content consumers and the constraints and requirements of the digital systems that manage it. Andy Fitzgerald shows how information architects use data, stories, and meaning to bridge these gaps and bring content to life for the human beings who need it.

https://ellessmedia.com/csi/andy-fitzgerald/

Content Strategy Insights interviews are hosted by Larry Swanson of Elless Media.

https://ellessmedia.com/csi/

A massive topic deserves a massive video. Rob Miles discusses ChatGPT and how it may not be dangerous, yet.

More from Rob Miles: http://bit.ly/Rob_Miles_YouTube

The 'Danish' Reddit post: https://www.reddit.com/r/GPT3/comments/zb4msc/speaking_to_chatgpt_in_perfect_danish_while_it/

Some of Rob's own videos which are relevant:

Reward Modelling: https://www.youtube.com/watch?v=PYylPRX6z4Q

Instrumental Convergence: https://www.youtube.com/watch?v=ZeecOKBus3Q

Problems with Language Models: https://www.youtube.com/watch?v=w65p_IIp6JY

https://www.facebook.com/computerphile

https://twitter.com/computer_phile

This video was filmed and edited by Sean Riley.

Computer Science at the University of Nottingham: https://bit.ly/nottscomputer

Computerphile is a sister project to Brady Haran's Numberphile. More at http://www.bradyharan.com

We have to talk about Twitter, right? Elon Musk bought it. He’s making all these changes, and he’s realizing that content moderation decisions are quite complicated, especially when the stakes are high.But talking about Twitter in a vacuum seems wrong. There are lots of other social networks and community-based products, and they all have basically the same problems: some technical (you have to run the service), some political (you have to comply with various laws and platform regulations around the world), and some social (you have to get millions of users to post for free while making sure what they post is good stuff and not bad stuff).So, we’re doing something a little different this week. First, I’m talking to Matt Mullenweg, who is the CEO of Automattic, which owns WordPress.com, the blog hosting platform, and Tumblr, the social network, which he purchased from Verizon in 2019. Then, Verge deputy editor Alex Heath and I are going to break down a bunch of what Matt told me and apply it to Twitter to see what we can learn.Okay, Matt Mullenweg, CEO of Tumblr, followed by Alex Heath. Here we go.Matt Mullenweg is the CEO of Automattic and the CEO of Tumblr. Welcome back to Decoder.It’s great to be back so soon. At some point you’re going to have to give me a badge or something.I think you are our first repeat CEO guest inside of a year. You were on in March, and we talked a lot about WordPress and a little bit about Tumblr. I wanted to have you back because you are one of the few people I know who has ever purchased a social network, and it seems like a really good time to talk about the challenges that come along with purchasing a large, at-scale social network with millions of passionate users. So welcome back.Before we talk about Tumblr, buying a social network, and obviously Twitter — which is the reason that talking about buying a social network is relevant — you are the CEO of Automattic, which makes Tumblr, WordPress, Pocket Casts, and WooCommerce, a very popular e-commerce buying solution. Just give the audience a really quick refresher on what Automattic is and how you think about the company.Automattic is a holding company that I founded 17 years ago to basically support the open web. We do WordPress.com, Tumblr, Pocket Casts, and Day One, which is a great journaling app. The new year is coming up, and a lot of people like to start journaling at the start of the year.Thank you. We are trying to make the web a better place with everything that we make. We’re always asking, “How can we put users more in control? How can we align our business model more with what our customers and users want?” Open-source is obviously at the core of everything we do. WordPress is open-source, GPL (General Public License). We also open-source Pocket Casts. We are open-sourcing Tumblr, but it is taking a while. Sorry, I think the last time I said it I was a little more optimistic. It’s just a lot of code, but we will get to it. “I believe open-source is a fundamental human right... It’s just as important as freedom of speech, freedom of religion, or any other freedom.”I believe open-source is a fundamental human right. As technology takes up more and more of our lives, it’s just as important as freedom of speech, freedom of religion, or any other freedom. It’s important to have the freedom to see how our software works and to modify it.Like I said, we did an entire episode with Matt about Automattic and all those ideas back in March. If you want to dive into that stuff with Matt, go back and listen to that episode. We are going to focus pretty narrowly on Tumblr today. Like you said, Automattic is a holding company. You are also the CEO of a couple of the companies inside of that holding company, notably WordPress.com and Tumblr. You only recently bought Tumblr a couple of years ago. When you purchased it, you ran into some issues — some of which are unique to Tumblr, and some of which, to me, just seem like the problems of owning a social network. Can you quickly go through that? You bought it from Verizon, which had come into owning Tumblr because Yahoo had bought Tumblr. Honestly, the parallels to Twitter seem striking to me.Listen to Decoder, a show hosted by The Verge’s Nilay Patel about big ideas — and other problems. Subscribe here!It’s also a roadmap. Since buying Tumblr, we have launched subscriptions and we are open-sourcing the algorithm. We’re doing a lot of things that I think are on the roadmap for Twitter too. I will say that it was probably the most humbling thing in my business career. We had been running WordPress for a while at that point, which powers a good chunk of all the websites on the internet, so I thought I had seen it all. Tumblr is a large-scale social network that is only a fraction of the size of Facebook, but we started encountering issues that were beyond my previous understanding of content moderation and free speech. We’re seeing that happening in real time at Twitter. They will allow someone like Ye back on, and then take him off again. There’s a phrase we use for a huge amount of speech, and that is “lawful but awful.” It might hurt people’s mental health, incite harm, or be really mean, like bullying, but it’s not technically illegal. We’re a private company, so I guess we could host it if we wanted to, but you need to think about your responsibility to society and to your users. It’s as if you were hosting a party. It behooves you to provide a safe environment for everyone there that includes food, water, restrooms, and all those sorts of things. I feel like when you’re hosting a social network, it’s your responsibility to provide a safe and healthy environment.I want to get all the way there. Honestly, I want to spend most of our conversation talking about what those kinds of decisions are like. I think far too often the casual observer thinks they’re easy, and far too often Elon Musk acts like they’re easy. In reality, anybody who has tried to make those decisions, like yourself, finds out they are extraordinarily complicated. The tradeoffs are bad and everyone is going to hate you after. But let’s start at the beginning. Why did you buy Tumblr?Tumblr was always WordPress’s best competitor. I feel like Tumblr combined the very best parts of blogging and social networks, and it innovated the form of social media by introducing multimodal posts. One, I was excited to bring some of the fun back to blogging, because I think that everyone should blog more.Two, I wanted to see if we could create a mainstream social media that wasn’t reliant on surveillance capitalism or advertising as its primary business model. We run ads on Tumblr, but we also have upgrades that turn off ads, and we’re introducing lots of other subscriptions — some fun, some serious. If we can make it a subscriber-supported thing, then we can truly be aligned. Even if I were no longer running Automattic or Tumblr, the business model would align the users with its business. Finally, I felt like we need a space on the internet for creativity, art, and artists. The other social spaces on the internet have gone different directions. Twitter became a lot more about arguing, Instagram became about showing off, and Facebook became about weird people you went to school with saying weird things. Tumblr always had this frisson, this magic. Instead of an angry mob, it’s more like comedy improv. There’s a “Yes, and…” to it.Tumblr is a collaborative group art project at scale, for sure.Totally. We have seen some amazing examples of that, even in the last few weeks with Goncharov. At its best, it’s like, “Well, what if people’s social media time could go to something like that?” It’s something that puts a little more control in the hands of users. You should feel good after using it and you feel creatively charged. That’s what we have been working on since we bought it.Again, I say there are parallels between the Twitter and Tumblr backstories. They start around the same time; they both have these young, famous entrepreneur CEOs; they have these crazy valuations; and they have these administrations and changes in control and ownership. It seemed like nobody really understood the product they accidentally made. Twitter is at a place now where Elon bought it for $44 billion, and he says he overpaid for it. Tumblr is slightly different. Yahoo bought Tumblr for $1.1 billion in 2013, and then there was a series of ownership changes. Then you, Automattic, bought it from Verizon for $3 million in 2019. How did you come to a price for a social network? We’re all talking about the $44 billion for Twitter. You actually evaluated a social network, its technology, and its user base, and you said, “You know what, this isn’t $1 billion, it’s not even $100 million, it’s $3 million.” How did you come to that price?“Tumblr was, and still is, burning quite a bit of cash.”Have you ever heard the phrase “free like a puppy”? The transaction cost for us buying Tumblr was de minimis. But it was a deal in which we took on all of its liabilities and all of its legal cases, we kept all the employees and all the costs to run it. Tumblr was, and still is, burning quite a bit of cash. People were like, “Oh, you could buy an apartment in New York for that,” but you would be buying something that costs $60 or $70 million a year to run. You would be taking on all of those obligations as well.There were definitely people who would have paid more for Tumblr, but to the credit of Verizon and CEO Hans Vestberg, they really cared about it going to a place where the community would be well-stewarded. I very much think about that. I’m the third CEO of Tumblr; it was David Karp, then Jeff D’Onofrio, then myself. I’m stewarding it for the next generation. I’m not going to be CEO of Tumblr forever, and I’m going to find someone to take it over at some point. But I hope that it’s around 20, 30, 40 years from now. As we can see, kids still need it — Tumblr’s user base is still primarily under 25. It’s this weird thing that fills a role on the internet that nothing else does.When you say it’s de minimis, you are basically saying that Verizon sold it to you for the smallest amount they could. And that you bought it for the smallest amount you would pay, knowing that its carrying costs were so high. Was that a straight conversation? “Hey, we’re not going to figure out how much money Tumblr makes and do a multiple of revenue to come to a valuation. This is bleeding cash, and it needs a good home. We’ll be that home. What’s the smallest number your board of directors will accept?”I think what Verizon cared about was the employees and the user base, which is also what Automattic cared about. We really oriented that. Internally, we budgeted about $100 million that we were going to spend on Tumblr to turn it around.Well, when you think about the burn and everything, that was our calculation for it. It wasn’t the $3 million, so people made a big deal about that number. In business, it’s too good to be true. You couldn’t buy something like that for that amount. You have probably seen this as well, where there are naval bases or missile silos that you could buy really cheaply, but you are then responsible for the cleanup. We had to do a lot of cleanup when we bought Tumblr. There was a lot of hate speech on the platform, and they were behind on content moderation. I think the technical infrastructure had started to degrade quite a bit. We’ve spent the first couple years really just rebuilding things.Automattic is a deep-tech infrastructure company, so we were able to bring it onto our infrastructure, rewrite a lot of things, make it faster, make it more stable, and also bring our experience and our values in terms of moderation. I took over as CEO in February, and we were just rounding that corner of all the cleanup work when all the Twitter stuff started happening and people started saying, “Well, maybe I need an alternative.” Particularly in the last month or so, we have seen some huge waves and droves of users. Fluent celebrities like Ryan Reynolds, Lynda Carter, and Halsey are coming over or coming back. It has been a fun time to be on Tumblr. I tell the team that fortune favors the prepared. There are also a lot of other places people could go, but we are ready for the waves. We can handle 200,000 to 300,000 sign ups a day. We can handle what’s happening.You bought a social network. You set a price; now you have it. You have to carry all the employees, and you have to re-architect the technical infrastructure, which Automattic is good at. You are good at it. Then there is this big moderation question. Let’s start with the employees though. The first thing Elon did was fire everybody. That didn’t seem like the first thing you wanted to do. Talk me through that. How did you evaluate what you had bought in terms of personnel?Well, I think the tough thing in an acquisition, and particularly a turnaround, is that you’re buying it for a reason. If it was doing well, it wouldn’t be a turnaround. Obviously, some of the existing employee base has not been as successful as they or you would have hoped, which is a nice way of saying some of them probably shouldn’t be there. But we also bought it because it was working. Tumblr still had this really vibrant user base, despite what I would say was corporate mismanagement and misalignment of incentives. It was still growing, and there were still a ton of mobile ties and young users, which was very, very interesting to us.How do you not throw the baby out with the bathwater? We brought the whole team over from day one, and we also tried to switch a bunch of people that were long-tenured Automatticians, which are people who have been at Automattic for a long time. I actually took some of my very best people in the company and switched them over to do different jobs inside of Tumblr — engineers, designers, et cetera. That helped us merge our cultures, identify low performance, and reshape the team for what was needed now. Tumblr also had a lot of attrition. Verizon shared a building with Facebook, so good Tumblr engineers were getting poached in the elevator. This was also the time when there was crazy tech comp. That has settled down now, but at the time it was a little wild. So we have remade the team, we have made the tech, and we’re starting to remake the product. I’m also very excited because now we’re starting to have some fun. You saw the blue checkmark thing. We’re also starting to innovate on the format a bit. On Tumblr, for example, you can now have a post which has a gallery and a video — it’s basically multimodal social media posting. Blogs have done this for a long time, but we’re bringing it into the social media form and onto mobile. That’s fun for me, because the creativity that is being expressed there is more than what you can do on any other social network right now.That timeline is really interesting. You bought it in 2019, but it’s not until the very end of 2022 when you’re saying, “Now we’re having fun.” That is a long time to integrate the cultures, reset the expectations, and then get to product innovation. Or maybe from a user perspective it’s a long time, but from your perspective maybe it’s lightning fast. Which one do you think it is?“This has been harder than anything I’ve done before.”I said it earlier and I’ll say it again: it’s the most humbling thing in my business career. I have been doing this a while. We have done successful acquisitions, like WooCommerce and other things, but this has been harder than anything I’ve done before, which is why I stepped in to run it directly in February. We weren’t seeing the amount of turnaround that we had hoped for.Do you think that part tracks with the Twitter timeline? Musk comes in, he takes over Twitter, he’s like, “screw it,” and then there’s a gigantic, sweeping culture reset and gigantic, sweeping public comments about how the company was trash at every level. Do you wish you had done something like that? Do you think that would have been effective?By the way, just to provide you cover for this answer, there’s a part of me as a leader that is sometimes like, “Maybe I should just run around saying everything is trash and reset.” There’s something in this way of working that I think every leader finds tempting. Most people are not this maniacal. I am not. I wouldn’t be able to sleep at night if I did this to my team.If you’re in any leadership position though, there is something appealing about saying, “Oh man, I wish I could just clear the deck.” When I was much younger and meaner, I worked at another large company — which shall remain nameless — and I would often think, “I wonder if I could just fire half the people on this floor? Would anyone even notice?” There’s something about a big company that engenders that kind of thinking. Now that I have provided you cover, do you think that you should have done something more drastic?I don’t know. It’s very hard to play it back again. About 85 percent of the team is new on Tumblr, or was not there at the acquisition. That is a pretty big switch that happened over a couple of years. Some of it was natural attrition, and some of it was performance management. A billion a year in debt service interest payments? That’s a pickle.I think that each company is unique and each situation is unique. For us, one of our concerns is definitely the burn that Tumblr has, how much cash it’s burning over what it makes in a month. The deal with Twitter is that he paid so much, set the price, the market crashed, then there’s a whole lawsuit that forced him to buy it, and now he has this huge debt load. It creates, I imagine, a whole other different set of constraints. What are they saying? A billion a year in debt service interest payments? That is tough when combined with advertisers leaving. That’s a pickle. I do not envy their position. Our position right now is actually interesting. As we’re approaching the end of this year — and I think I said this in March — we are at a pivot point where we’re like, “Well, do we need to make Tumblr into a smaller team?” That, for us, would not mean layoffs, but just switching people to other divisions in Automattic, all of which are profitable. So that’s saying, “Hey, some of these folks go work on WooCommerce, some people can go work on our enterprise VIP, and some can work on WordPress.com,” and taking Tumblr down to 50 or 60 people, which the business could support. Right now, it can’t support 200. Actually, thanks to the Twitter stuff, all of our investors are now like, “Wow, you look really smart for doing Tumblr. You spent way less and the metrics are looking really good.” We have definitely had waves of users and advertisers coming over saying, “Hey, we want an alternative. We disagree with some principles or things that Musk has said or done. Can we come spend money with you?” Which is great. There is also something that I’ve just never seen in my business career, which is the talent exodus. As an example, Stripe laid off 15 to 16 percent of its staff, which was surprising, because Stripe is a golden child. You assume that they tried to do that as performance management. They tried to take off the folks who they considered to be low performers. I would guess that Twitter was probably going to do a layoff anyway, though probably not as big as it was nor as sloppy. I’m sure they were going to, because every other tech company has. Now as an example, Stripe will make mistakes there. You could assume that some percentage of the people they laid off, who they considered to be low performers, were actually quite good, but you would maybe be cautious to hire out of that layoff. If we were targeting that, we would assume that these were not the folks that had Stripe saying, “These are the crucial people we need to keep.” With Twitter, you have lifers in both the initial cuts and with those who left afterwards because they disagreed with some of the public statements — like the Paul Pelosi tweet, or whatever it was. You hear people talk about missionaries versus mercenaries in companies. There are the folks at companies who are core to the culture; they created the tech and they know it in and out. You never let those people go. You do whatever you can to keep them. Perhaps as part of the fog of war in that Twitter acquisition, knowing that they had to make these big cuts, they let some of those people go, which then also caused other people to leave.“I would be happy to hire 50 or 100 people who left Twitter. We have publicly said we’ll hire even entire teams.”Every single other tech company, including us, who said, “We’re slowing down hiring,” has really reached out. I mean, I would be happy to hire 50 or 100 people who left Twitter. We have publicly said we’ll hire even entire teams. Those conversations have been quite intense over the past few weeks, and they include some executives who are helping us navigate the 5,000 people who left. We’re now in the mode of, “Well, let’s see if we could create an amazing machine learning AI team inside Automattic with some of the folks who left Twitter.” It has been kind of an odd shift. Again, I’ve never seen anything like that in my business career.It’s funny. I feel like you and I have come up after those moments in tech history. Intel was created because a team of people left Fairchild Semiconductor, so now Intel exists and we think of it as an institution. It feels like maybe we’re living through that moment again as tech companies have these gigantic layoffs, where entire teams of people who like working together are now available. The Twitter Spaces audio team that built Spaces is just like, “We’ll come do this for you again, but we want to work together.” Would you build a live audio product? Would you just go hire that team?I don’t know if we would do a live audio product, but I think that team is quite good. We put up a dedicated landing page which was basically a distillation of a lot of the conversations I had. The first line on the page is, “We love Twitter,” which is true. I also love Twitter. If you’re hiring the missionaries, you don’t do it with a “we’re going to crush Twitter” message, because they have poured their heart and soul into Twitter over the past however-many years. They are not motivated to kill Twitter, because they also love it. What I think is interesting is asking, “Hey, could we do it again, avoid some of the mistakes, and create an alternative?” To me, the best way to actually influence Twitter now is by creating great competition from the outside. I think Twitter will survive and will still be around 20 years from now, but I think it will be made better if Tumblr is there nipping at its heels with some really excellent user experiences, maybe innovating the forms, and just pushing the bleeding edge. I mean, maybe Tumblr will always be the smaller but more innovative network. That’s entirely possible. That is the kind of space it was before. Cool. Guess what? We’ll make the rest of the web better. “You need a good nemesis and competitor in business, otherwise you get lazy.”That is also WordPress’s thing. I want to create a web that’s fully open-source. There are a lot of great proprietary competitors, like Shopify, Wix, Squarespace, et cetera. We forced them to open up by being the open alternative, much like Android kind of forces iOS to unclench a little bit and open up some of the things they do, like not allowing changes to defaults. Android forces iOS to be better. You need a good nemesis and competitor in business, otherwise you get lazy. A few quick questions, just to complete the acquisition story. You said the burn when you bought it was $60 million-ish a year. Are you closer to profitability now or is that burn the same number? No. We brought it down, but we would need to grow Tumblr’s revenue by another $20 or $30 million a year to get it to break even at its current people cost. And you’re thinking of hiring more people from Twitter?And we’re hiring more people. Now, the good news is that one thing we’re starting to do is combine some of the teams. For example, Tumblr doesn’t need its own separate trust and safety, or terms of service. It did have one, but we actually have similar problems across all of our properties, like protecting against illegal content, responding to DMCAs quickly, and taking down hate speech. Those are similar issues, so we’re able to use some of the same backend tools to monitor every upload and other things. It’s not only a Tumblr cost at that point. I think between WordPress.com and Tumblr, we can actually share a lot of the backend infrastructure. As we’ve said, we want to switch Tumblr to actually be powered by WordPress. I think of it as two great restaurants that share the same kitchen. The restaurants have totally different vibes, different user interfaces, different menus, but they’ll have the same excellence and ingredients — the same backend, essentially. I think that’s very, very possible.This is the food hall model for an infrastructure company.That’s very good. We have talked a bunch about hiring and size, and you said you’ve had 85 percent turnover. Is the Tumblr team bigger or smaller than when you acquired it?It’s probably a little bit bigger or about the same size. I think it will probably get about 10 percent bigger in the next six months.This is one of the classic Decoder questions. How is that team structured? Is it structured the same way as when you acquired it, or have you reallocated some of those numbers?I mean, I have come in as CEO, and naturally the big changes to the structure are because I’m running a lot of different things, including Automattic and WordPress.com, separately. I very much have a leadership style that pushes a lot of things onto the leads within the company. I’m not like, “We need to have a meeting every day with the executive team and blah, blah, blah.” I’m more like, “Here are the five most important things to me, get them done. What are your most important things? What am I missing?” Automattic does a lot of asynchronous communication. There haven’t been big changes. I have brought over some folks from Automattic to help me out, including a chief operating officer and others, that are really helping with the day-to-day. I would say it’s a fairly standard structure other than that.I have a thesis that the actual central product of a social network is content moderation. Every piece of the puzzle of the social network is trying to incentivize the people who create for you to make things that are good that you want, and then trying to show the user, the audience, the best stuff on the social network. Obviously, what comes along for that ride is, “When people post bad stuff, I want it to go away. I want to disincentivize the bad stuff in potentially harsh ways, and I definitely don’t want to show it to an audience.” That’s the part we almost always focus on. Then there’s the other set of incentives. “Here’s what we want from you and here’s what we want to show people.” The positive incentives almost always go completely unrecognized as part of the content moderation puzzle. Does that seem right to you? Is it fair to be that reductive and say the product people at social networks are basically making content moderation?Yeah. What was your post on this? What was the title of it?Oh yeah, I think I tweeted that. I don’t agree with the title headline, although it was a good one. But man, it was spot-on. You really nailed it. Everyone who is listening to this, go read that post, because I think you hint at and also link to some of the subtleties of content moderation and doing it at scale. When you think about it, the number of Tumblr’s weekly active users is larger than any US city. Everything that happens in a city in a week can happen on a social network, even though we’re one-tenth the size of Twitter and one-hundredth the size of Facebook. People don’t anticipate that. You also see all the new social networks — the Parlers, the Gabs, the Substacks, and even the Posts or whatever — struggle with this early on.There is a learning curve. Even if you hire people, even if you know what’s going to happen, whatever user base you attract is going to cause new types of problems. For example, Tumblr has a younger demographic, teenagers, that are maybe stereotypically a little more angsty, so mental health things are really big there. We build a lot of stuff in so that if you search for certain tags, before we show you, you will get a message that says, “Hey, do you need help? Here’s a phone number. It gets better.”There was one thing that I learned about called the pro-ana community, ana being short for anorexia. This was a community of folks who were using a social network in a way that’s not illegal, but was basically encouraging anorexic behaviors. I’m not an expert in this at all, but my understanding is that it is a mental health challenge and ultimately quite physically debilitating for people who suffer from this. If you’re hosting and promoting content that’s encouraging that, what are you doing to those kids, those people, as a society? Again, it’s not illegal, but it is your responsibility to control the distribution of that, to tamp it down if people are posting it, and to try to provide them pointers to resources — because we’re a tech company, we can’t help with that. There are lots of nonprofits and people we can point to that are actually professionals at this, and we can try to nudge people in the right direction.By the way, that really works. There are some untold stories where tech has made society undeniably better. I’ll talk about two issues. One that has been covered a little bit is around child exploitation material — people who abuse children, take pictures, et cetera. Tech companies have basically come together and created technological solutions and data sharing that have become quite good at catching this. Then it all gets passed to law enforcement and they do their thing. I think that has helped quite a bit. Then there’s suicide prevention. On pretty much every social network and search engine, if you type in certain terms, they will jump in and say, “Hey, here’s pointers to resources.” There has been a lot of sharing on what people click more, what resources are best, how to provide a phone number, how to do this internationally, how to do this in every language, et cetera. I think tech companies, including competitors, share this quite freely with each other, because we all agree this is something that is part of our responsibility to society.Those are things where Automattic as a company has a set of internal values, and you are shaping the product in line with those values. Those things are horrible, I don’t mean to diminish them at all, but preventing them is universally agreed upon, right?Yeah, it’s not controversial.Right. We should not encourage young people, especially young women, to be anorexic. That’s a huge problem in that community. And we should aggressively intervene in suicidal ideation or communication to stop it and provide resources. Those are aggressive moves. You’re saying, “We’re going to stop the speech, we’re going to shut it down, we’re going to show up when we see this stuff to get in your way and say, ‘Go to these resources.’” They are aggressive interventions, but they’re not controversial. Then there is a universe of stuff that is totally controversial, where even the slightest intervention gets you in hot water. The one that I think is easiest to point to in regards to Tumblr is porn. There’s a lot of art on Tumblr. It’s famous for being an artistic community. There’s also a lot of nudity and a lot of straight-up porn. At least, there was before.I think Verizon as the owner, being a telecom company, was like, “No, this isn’t us.” Then there’s Tumblr itself, which has some values that might support pushing the boundaries of allowed speech past what a platform like Facebook or Instagram might allow. Then there’s Apple, which says, “We won’t let this app on the store if it has content that we object to.” Then there are credit card processors, like Visa and others, saying, “We won’t support transactions for pornography.” The lines aren’t even clear on what is porn or what isn’t, or on who is qualified to say what people on platforms should be looking at besides the platform itself. The values inside the company might not line up with what you want users to be able to do on the platform. There are 8 million external actors with their own values that have influence and existential control over the platform. Walk me through this. I’m picking on porn, but I can pick any number of other speech areas that have the same exact problems.“‘Go nuts, show nuts’ was actually Tumblr’s previous policy on adult content.”I’ll give you the TL;DR. I wrote a post about this and it was titled, “Why ‘Go Nuts, Show Nuts’ Doesn’t Work in 2022.” So “go nuts, show nuts” was actually Tumblr’s previous policy on adult content.That’s very Tumblr, like the most Tumblr thing possible.It’s very Tumblr. I think Tumblr had content moderation issues. Part of why they got shut down by Apple is that they were not doing a good job policing illegal content, in addition to the porn stuff. I wasn’t there at the time, but my guess is that Apple also wanted to make an example by shutting Tumblr down and removing it from the App Store, even though it’s owned by Verizon, which is one of Apple’s largest partners in the world. That must have really woken everyone up, like, “Hey, they’re taking this seriously.” The reaction was heavy-handed. Verizon is a very conservative company that has better things to deal with, like their hundreds of billions of dollars in revenue. They started manually reviewing every single post and reblog on Tumblr with content moderation farms in other countries. They were also applying what I would call very faulty machine learning. Famously, someone posted a picture of their manicure — they got their nails painted and posted a picture of their hand — and this got their account locked for being adult content. What the algorithm was looking at was, “Well, there’s a lot of skin tone as a percentage of this image, so this is likely adult content.” Now, combine that with one more thing: when we purchased Tumblr, they had a six-month backlog on support tickets. Imagine you’re an active Tumblr user, you post a picture of your manicure, and your account gets locked — in a really mean way, by the way. It was like, “You have violated blah, blah, blah. You’re locked out.” You appeal it and say, “Hey, no, this was just a picture of my hand,” but you don’t hear back for months. What a perfect formula to destroy a user base. That was where it was.Something we changed since our last podcast was that we actually reopened more adult content, specifically what we call “artistic expressions of the human form.” If you had posted literally Michelangelo’s statue of David on Tumblr before, the content moderation rules would have locked the post or locked your account. We got good at appeals and everything like that, but we were stuck with these old rules, and we couldn’t really change those rules until we had some better community moderation in place.We introduced a rating system where users can self-tag when they post something. “This contains X, Y, Z,” and X, Y, Z could be drugs, it could be violence, it could be the human form in adult ways. We still don’t allow what we jokingly internally call “things going into things,” or what people might call hardcore pornography. Something like that is still not appropriate for the service, but there is a wider aperture or Overton window for what’s allowed, which actually matches what we’ve done at Automattic for a long time. We were kind of unifying Tumblr’s position with ours. It’s interesting, because Elon also talked about bringing the MPAA movie rating system into this, which is actually where we started in the first iteration of this feature. When you get into the history of the rating system though, it’s actually quite fraught. Just think about it. If there was one female nipple in a movie, all of a sudden it’s like PG-13 or R, but then there can be any amount of violence, gore, and blood spurting out — which obviously is not great for kids either — and that could be rated PG. We went to a form of classification, a taxonomy, that was a little more nuanced. So if you sign up for Tumblr, all that stuff is off by default. You can’t turn it on inside the app because of Apple, but if you go to web, there are toggles where you can say either, “I would like to see this stuff normally,” or, “If this loads on my feed, I would like it blurred out by default.”Yeah. You open your Tumblr and you’re browsing through, you don’t want that stuff popping up when someone walks by. That’s embarrassing for everyone involved. We really thought about it from a user-centric point of view. We have seen that this actually aligns incentives. Let’s say you’re a burlesque performer in New York City. Bathtub Gin, right? It’s an awesome, famous, burlesque place. You want to post pictures from your performance. These are mature, you don’t want kids to see these, so you can tag this and now you know that it will be protected. Folks under 18 won’t even know it exists, but people who want to see this can find it. Everyone’s happy. The incentives are very aligned. The violations now are not for what you post, but for mistagging. We take mistagging very seriously, because, obviously, that’s wrong. It could endanger kids. It could do lots of things. If you’re tagged correctly, we allow you to post a lot more stuff. We have done this while navigating Apple’s App Store, the credit card processors, and everything else.If you flip on the toggles on the web, does that stuff show up in your feed on the iOS app?Does that comply with Apple’s rules?Yeah. Well, how do Twitter and Reddit get away with it? They allow everything, like “things going into things.” Pretty much anything you could find on a porn site is also on Twitter and Reddit. How do they get away with it? One, maybe they’re just too big and they have enough legit content that Apple wasn’t really worried about it. Two, maybe they also made these web-only toggles. We decided to just copy that feature.Did you have to go to meetings with the App Store review committees, with Phil Schiller or Tim Cook, to get this stuff through? Or did you just submit the app and hope?No. No, I’m not that fancy. I would love to meet Tim Cook actually, but I never have.Well apparently you just have to tweet at him and sic a bunch of Republican congress people at Apple and an anti-trust bill and you get a meeting.“The App Store review process is still a black box and still capricious. You never know what’s going to happen.”That works for Elon but I don’t know if it would work for me. The App Store review process is still a black box and still capricious. You never know what’s going to happen. We’re trying to launch this new feature for Tumblr. I would love to announce it right now. It’s going to launch any day now. It has been ready for weeks, but the in-app purchase was denied. You get this weird thing where versions of the app are not approved, but it’s a feature which is not controversial. It’s totally by the book for using in-app purchases. We’re not trying to skirt anything, yet somehow it’s been held up for weeks with back and forth on various content things. I think we made a mistake in submitting the app one time where we set a toggle wrong, so that then creates another week. Then Thanksgiving happened. It’s an odd platform. Most of our tech, you could just ship whenever you want. You can test, you can put things up, you can take them down. In the app stores, it goes through a person, and depending on who the person is, they might interpret the rules differently.This is actually a thing I missed in the “Welcome to Hell” article. I talked about how these decisions are hard. They’re political decisions, not technology decisions. Then you have governments. The speech laws in Germany are very different from the speech laws in the United States. Here in the United States, Florida and Texas have social media moderation laws that are government speech regulations and they are probably in violation of the First Amendment. It’s very complicated just as a political actor. Then as a tech company, your primary distribution platform is controlled by two companies that more or less operate in lockstep when it comes to moderation issues, and they’re completely opaque. It’s the thing I missed because it’s maybe the wonkiest thing, but it strikes me as, “Oh boy, I screwed up.”I’ll disagree with you there.I would say Apple’s and Google’s app store moderation is night and day. With Google, you get awesome tools, where you can roll out to percents of users and then roll it back. Everything is really fast and they allow way more stuff. They’re not as draconian about forcing in-app purchases. It’s totally different. Apple is the most powerful player in the market, especially in the US. They are a monopoly. They control everything. They’re also opinionated. My interpretation of why Apple is so strict about these things is they take their responsibility to their users quite seriously. There are examples of this.If you sign up for a New York Times subscription on The New York Times’ website, they make it really hard to cancel it. You have to chat with someone and it takes 30 minutes of your time. It’s like canceling a gym membership. It’s terrible. It’s a horrible user experience. If you subscribe to The New York Times through Apple though, you can just click a button to cancel your subscription. I think that is Apple advocating on behalf of users for something that is user-friendly. Now, they have things we probably all agree on, like canceling subscriptions, and they have a section that they do. It feels like they still think they’re the underdog.It’s like they still think they have an existential risk of being snuffed out any moment. I’m excited actually, because Apple has more cash in their bank than most countries. They are one of the most powerful entities on the planet, even more than most governments. I’m seeing them starting to shift into more of a benevolent role and realizing their size and their power.Well, I think they’re getting pushed into shifting by regulators and governments.They’re being pushed, which is great. The EU and the US are all starting to push a little bit. I bet internally, as well, there are folks inside of Apple who would agree with my position on these things. It also might be generational. As new leaders come up through the organization, perhaps that will shift some of their policies.What you’re describing are people who didn’t come up as the underdogs. Most of the current executives were there when Apple was the underdog. You just see their culture over time. You have described taking over Tumblr as the most humbling experience. It’s this stuff. “Now I’m the politician who is in charge of a large city or a small country. The users are doing whatever they want, and all I can do is incentivize them to do good things and not bad things. There’s also a host of other constituents — app stores, credit card processors, whoever — who are deeply interested in whatever I do.” Where do you have authority to make decisions, and what do you think the limits are in that authority? When you say humbling, it seems like that’s at the heart of it. You’re not a tech executive who’s saying, “Make the button blue.” You’re a politician who’s saying, “I hope I’m going to make a policy decision that is expressed out through all these constituents and will achieve the result I want.”I guess the best way to probably summarize where I am in 2022 — and this has evolved over the past year and the past 10 years — is that I’m extremely libertarian in terms of what people should be allowed to say. By the way, I’m totally okay with things I disagree with strongly or with people saying bad things about me. I’m a public figure. Great. Where I think I have become more conservative is in bullying and hate speech, that sort of stuff. Of course, calls to violence are pretty noncontroversial. I would say bullying, or trolling, is maybe more in the middle. If you remember Flickr in the early days — like with Stewart Butterfield, Caterina Fake, Heather Champ, and Derek Powazek — they fostered an amazing community, often manually, by going and commenting on new users or choosing what they highlighted.We are trying to bring some of that idea of community building to Tumblr. I would say that it is about 20 percent pruning out the bad stuff as if you’re weeding a garden and about 80 percent encouraging the things that you want to grow. It definitely needs to be a long-term thing. You need to water it every day, but the results are going to happen over months or years. That is where we’re at right now with Tumblr. Take something like the Goncharov thing. People should Google it. It’s a Martin Scorsese film from 1973.That whole fun, amazing, beautiful thing happened partially because we created a space where you could have a “yes, and…” improv environment, with people riffing off each other and without a few bad actors coming in and spoiling it. I think we’ll see a lot more stuff like that on Tumblr in the future. It actually keeps growing, too. What are those posters called that they put up in New York? I saw a picture yesterday, and there’s actually Goncharov posters now.Yeah, on walls in cities. It just keeps going.This is one of the things that I think is really interesting. The Tumblr community is going to be very excited that the CEO of Tumblr is deeply aware of this… I don’t know how to describe it. It’s like an improv art performance of a Martin Scorsese movie that may or may not exist. One of the things that people are excited about with Elon taking over Twitter is that the man is addicted to Twitter. It’s very obvious, right? He’s on Twitter all day long.He’s goofing. He’s posting memes. You’re obviously deeply aware of Tumblr and the community. You’re in it. Do you think it’s important for you, as the leader, to be consuming the service as a member of the audience? Because I think it cuts both ways.One hundred percent, yeah. There is a little bit where I do understand. My guess is that folks, like a Mark Zuckerberg, a Parag Agrawal [former Twitter CEO], or other leaders of social media, use the platforms a lot, but probably just under a secret account. They have an alt.You also need to be sensitive. My preferences are not the preferences I’m imposing on the entire community. I’m super liberal, all those sorts of things. That’s me. I’m going to be open about that. I’m also not saying people who disagree with me aren’t welcome.Tumblr is smaller than everyone else, but they’re still at scale compared to most everything else. When you think about running a social network at scale, it seems like we over-rank the part where there’s a liberal bias inside the content moderation team that we must overcome. We then significantly under-rank that people don’t want to be in a platform full of racists. Very importantly, advertisers don’t want to be near that content, and that’s causing the content moderation decisions. Has that been your experience? Do the advertisers exert pressure on what things you allow?I think there are two levels to this. One is overt pressure. This is advertisers saying, “I disagree with xyz,” and they leave. They vote with their wallets, which they’re welcome to do. It’s a free market. It’s capitalism. That’s kind of the expression of it.If you don’t buy ads on Twitter, you don’t support free speech. That’s what I’ve been told.I don’t know if I agree with that name and shame. I would call that more capitalist activism, which I think it behooves all of us to do. We should vote with our wallets and try to support companies that agree with our principles, not spend money with those who don’t. There’s a second level though that I think is just inherent to the business model, which I talked about with surveillance capitalism earlier. Sorry, I’m blanking on the name of the author who wrote the book on this.Ah, thank you. There is something inherent to what advertising fundamentally is, which is the business of influence. “I want to create a desire in you. Maybe you’re perfectly happy and content with your life, but I’m going to create an emptiness in your life — a want or a desire — and you are going to fill that with my toothpaste or my headphones.” That’s fundamentally what advertising does. The intersection of that and democracy is something I think we’re grappling with. If democracy says that free, informed citizens are able to vote on people and vote on how they’re governed, I like that model. There’s a social contract and a principle morality to it that we can all agree to as participants in the system, which I think social networks and private companies miss. You don’t necessarily vote for the policies and elect the leaders of Facebook.As personalization, targeting, and machine learning and AI become so good, technology’s ability to influence you becomes amazing. We’re seeing this today. How good is the TikTok algorithm? How good are Instagram ads? “Gosh, they know me so well. I buy more stuff off Instagram than any other place. They’ve got me dialed in.” That being applied to political influence is playing off in both sanctioned and unsanctioned ways. Actors like China, Iran, Russia are taking advantage of our free and open society to influence Americans. It’s the whole thing. “We were all worried about hacking the voting machines because that was a good story, but it’s way easier to just hack the people and influence the voters.”We were all worried about hacking the voting machines because that was a good story, but it’s way easier to just hack the people and influence the voters. The voting machines are fine, just influence the voters. It happens in every election. We know this for a fact. It’s not a conspiracy. How do we protect and inoculate society against that, when the business models of these networks are designed around the engagements and the influence, essentially?Has this been an issue for you with Tumblr? That you need to serve advertisers? Obviously, they are the money. You’ve only just rolled out some creator monetization tools, but advertisers are still money. Are those revenues growing, or are you saying, “Gosh, this is kind of icky, we need to get away from this”? You said a few minutes ago that advertisers are starting to come to you because they’re leaving other platforms. Do you find yourself trying to navigate that balance?We’re trying to balance it. I think that if you provide a free service, advertising is the only business model. Running a social network is incredibly expensive. When you sign up for web hosting, you pay money, you get a certain amount of space and a certain amount of bandwidth, and there is a hard cost. When you sign up for a social network, you can upload unlimited video which can be viewed an unlimited amount of times, and it’s essentially an all-you-can-eat-for-nothing plan.The companies still have to pay those bills, though. They have to build the data centers, they have to pay for the network, they have to do all that stuff. There is a real cost associated with it, and advertisers subsidize that. What we’re trying to do is create a model where half or more than half of Tumblr’s revenue is from subscribers. I think that gives us the ability to not be unduly influenced by advertisers. There’s not as much of an incentive to tune the algorithms in ways that create the engagements, enrangements, and loops that will influence and emotionally charge people. If you’re worked up, you’re in a state that’s more receptive to changing your toothpaste brand. That’s science. That’s fact. That’s human psychology.  It’s our lizard brain.We want to create a space which is much more creatively charged. We want Tumblr to be like going to a music show or a museum. You’re going to see some stuff that you haven’t seen before, you’re going to discover new stuff, and you’re also going to leave creatively charged. That’s not a mindset which is as conducive to advertising, but I do think that we can find a set of advertisers and a set of products that fit well with that. Again, Tumblr’s biggest benefit right now is that it has no golden handcuffs. As we’re creating an advertising system, we can start a little bit from zero, so everything new is good. I can’t imagine the struggle of having billions of dollars of revenue and trying to shift your advertiser base or your policies.Especially when the advertisers are saying you have to keep moderating as much as you have been, and your entire stated purpose of buying the thing is to moderate less. There’s a tension there that I think is really difficult. As the CEO of Tumblr who is trying to build an advertising business, do you explicitly hear from big companies, “We need you to measure your brand safety before we show up and give you money”?“I would say we can unlock a lot of revenue, but we have decided to not do the tracking and targeting that everyone else does.”I would say Tumblr’s struggle with advertisers is actually lack of targeting. Some people might opt out of Tumblr because they’re uncomfortable with the younger user base and the kind of silliness of it. That’s fine. I would say we can unlock a lot of revenue, but we have decided to not do the tracking and targeting that everyone else does. That means it’s more of an uphill battle to get advertisers to spend money. We’re introducing some. There is some tracking I’m totally okay with, like device and country. That sort of stuff is very natural. A bit of it is both users and advertisers. There’s some that is actually quite enlightening. Your user base could do it or your listeners could go in and buy an ad on Facebook or Twitter. They all have self-serve tools. The amount of targeting you can do is kind of insane.By the way, for all the stuff the tech companies do, the telecom companies are way worse. With Comcast, you can effectively target a set of three or four houses and serve cable ads to just them, and then track that. Credit card companies and banks all share your financial data, and they’ll then correlate that with whether you spent money in the store. The amount of tracking is insane. The amount of geo data that gets shared is where we need governance to actually step in, because capitalism is not self-regulating well there.You’ve been talking a lot about the general problems. There’s a set of problems that are Tumblr problems that are very focused on moderation. Then there are the general problems of, “Okay, Apple is our distribution funnel and they’re opaque. They feel like they have a huge responsibility to their users that sometimes gets in our way.” Then there’s, “What is our business model?” Many companies are going through it. To wrap this up, I want to ask where you specifically think moderation belongs in the stack. I’ll draw the distinction between WordPress.com, WordPress VIP that you have for enterprise customers where you host WordPress for them, and Tumblr. There is this idea that the closer you are to the pipes of the internet, the less moderating you should do. So Comcast and AT&T should not look at the bits that are going across their network. Cloudflare maybe shouldn’t, right? They are an infrastructure provider that rides on top of those rails. AWS has a set of policies, like they won’t host white supremacist sites, but that’s basically it. That is the whole line.Then you get all the way up to, “Tumblr should directly intervene when people are encouraging anorexia,” which is way different than what you think about Comcast. Do you think WordPress is at a different layer of that stack? Is it easier? WordPress sites are just sites. You can say, “There’s some stuff we won’t do, but on your WordPress site, you can do whatever you want. On Tumblr, which is a service we run — that we monetize directly with advertisers and everybody else — we have to turn the screws even tighter.”Yeah. I think you summarized it really well. Ben Thompson of Stratechery wrote “A Framework for Moderation,” where he laid this out really well. It is true that at the base layers, you want to defer more to governments when it comes to what should be allowed or not, rather than the companies making arbitrary, unilateral decisions. That’s because governments have their checks and balances. We have courts, we have elections, we have all these things that say we should have a feedback mechanism as part of society, for these rules and for what should be allowed to exist or not.WordPress exists in all layers, so I would say Tumblr does, too. So what should you be allowed to post? That, again, I’m pretty open to. If it’s allowed in the laws of the country, sure, let’s allow it, even if I would disagree with it or consider it morally odious. We talk about freedom of speech versus freedom of reach. Freedom of reach is like, are you providing distribution to it? Am I surfacing your post on our search pages? Am I surfacing it in the feed? Am I algorithmically providing distribution to it? I think any company which is doing that has to be more opinionated in the moderation stack. As long as governments abdicate their responsibility here, we’re probably going to disagree some or all of the time with the decisions that companies make.Do you think the United States government should get involved, though? That seems to me like everyone wants someone else to solve this problem or make these decisions. The most likely set of actors that would do that are government officials, and they shouldn’t. Especially in this country, they should not make those rules. The First Amendment says, “Do not make speech regulations.” So I’m just like, my frustration…It doesn’t say that, actually.It does, right? I don’t think the government should step in and make rules about content moderation.The First Amendment is the most widely misunderstood.Okay, but I think this is the disconnect between basically everyone. Your position on the First Amendment is that the government should make some rules.Well, in fact, it does. The common example is yelling “fire” in a crowded theater in a way that creates an unsafe place.That’s a horrible example. That’s not a real rule.That’s a line of dicta from a case that was overturned. Everyone points at it, but there’s no law against shouting “fire” in a crowded theater.If you are creating harm, there are laws around voluntary and involuntary manslaughter. There are laws around hate speech. We have laws around certain types of crimes.There are no laws against hate speech in this country.If there’s a crime done with a hate element, it has different sentencing than if it was done without it.Sure, yeah. Now I’m with you, but you’ve gotten all the way to “you murdered someone.” There’s no law against hate speech. You can just say that people of other races are bad. You can just do it.There are laws in certain countries, too — in Germany, for example.Sure, in other countries. But in this country here, most people are like, “The government should make some rules,” and almost every example that I’m given is like yelling “fire” in a crowded theater or whatever, which is not actually a rule. The law in this country, the case that overturned the yelling of “fire” in a crowded theater, Brandenburg v. Ohio, changed the standard from, “You’re going to cause clear and present danger,” to, “Your speech is directed at inciting or producing imminent lawless action and is likely to incite or produce such action.” You can shout “fire” in a crowded theater all you want, you are not going to cause imminent lawless action. You’re just going to get people to get the hell out. So that’s what I mean.If people were hurt in that evacuation…I think there could be liability for whoever did that.This is what I mean. That’s liability for the actual harm. But it’s not the speech that did it.I get what you’re saying. People also misunderstand that the First Amendment doesn’t apply to us as a company because we’re a private entity. We can choose who we do business with or not. That’s one of the most common misunderstandings. You’re quite good at clarifying that whenever it comes up. So why do I think the government should be more involved? Because of the feedback mechanisms and the checks and balances.Germany is a good example. Germany, as a society, has decided that because of their history, they will take a firmer stance against Nazi-type stuff than America does — which is kind of funny when you think about it. Great, as a society they have decided that. That might change over time. The US has had tons of horrible laws in its history, maybe more bad ones than good ones. Those will evolve over time. Perhaps even the First Amendment needs to evolve over time, but how would we change that? It would require a new amendment, which would require states to ratify. There’s a really high bar for changing these things.Content moderation boards are essentially trying to recreate government in the private sector, which lacks accountability, feedback mechanisms, and courts.By the way, that’s a good thing. I think sometimes the slowness of government can be an advantage, because hopefully that deliberation helps forge a better outcome. It’s not a good example for that right now, with the polarization and the way the parties fight, but ideally, they reach a middle. Companies don’t do that. If you look at content moderation boards and everything companies try to do, they are essentially trying to recreate government a little bit in a private sector, which lacks accountability, lacks feedback mechanisms, and lacks courts. So it’s a weird system.So yes, I kind of do wish that governments had clearer and better laws around this. I also agree that when they have tried to wade into this, there have been some terrible outcomes, like FOSTA/SESTA. There are a lot of terrible laws that have come out of the government trying to regulate this stuff, but I remain hopeful as new generations of leaders come up. They are digital natives. Gosh, Beto O’Rourke used to be a hacker. He was part of Cult of the Dead Cow. I spoke to him, and he was like, “I used to be a web designer.”He can’t win an election to save his life, though.I know, but there are others who are coming up who can. We definitely have an issue where people are holding onto power for a really long time, since the ‘70s or ‘80s. We don’t have anything quite like this gerontocracy in history. As that starts to shift, I think that we’ll start to see a more dynamic republic. At least that’s what I’m hoping for. That’s who I’m donating to and who I’m voting for. As a citizen, I’m trying to advocate for more of that.I think this is as good of a summary of the problem as we can get to. You start with, “I bought Tumblr for the smallest amount of money that Verizon would sell it to me for,” and you end up with, “I’m hopeful for a more dynamic republic because we need to reform the speech laws of the country.” When you say it’s the most humbling experience of your career, that seems like the journey. I’ve heard this from so many tech executives who run networks like this. They get to a place where they want a more accountable external force to give cover for moderation decisions because the pressure is so high, and the only actor they can think of to do that is the government. Then you run head first into the First Amendment and say, “Okay, I wish we had some leaders.”At least speaking for myself, it’s not that I want to be removed from the pressure or the responsibility. I think that the responsibility and power put on me and our team is beyond what is warranted by the social contract in our society or from our users. I’m just philosophically saying that there’s a better system for this.Can you think of one that is not government speech regulations?If I did, I’d advocate for it. Can you? You’re a deep thinker on these things. You’re probably one of the best writers in the world on this.Yeah, and I’m terrified of government speech regulations. It’s my open bias as a journalist. I think they’re bad on their face. I see them in places where they work. Everyone brings up Germany, which has this long, tortured history. They’re still complicated and difficult in that country. I think from our perspective, thousands of miles away, we’re like, “That seems pretty good.” From the perspective of many people in Germany, this is more complicated than you think.Then there’s other countries, like India, where the government is like, “We’ll put you in jail if you criticize us,” and that’s just the end of that road. I think there’s more danger there for our version of democracy than there is benefit. I hear you on pressure and power and all these things, but I think it would be better — and maybe this is where we should really end — if all our companies were more competitive and what they were competing for in a vibrant marketplace was experiences based on moderation. If we all just admitted what social networks make is content moderation, both in terms of recommendations and creator tools that would incentivize you to make something.TikTok does not incentivize you to make text posts. It does not want them on its platform. It incentivizes videos. There’s a lot of them, but they’re hacks, which is fascinating to think about. The platform itself is not geared to make you post text. It’s geared to make you post videos. That is a content moderation, I think. That the users have done something else is just a fascinating dynamic inside of that platform. Tumblr is incentivized to make you post text and images and blogs. I think a blog post is a little work of art. I’m like you, I’m a blogger. Tumblr incentivizes you to make that thing. If we would just agree that the platform should be transparent about how they are moderating and what the rules are, we would stop yelling inanely about the First Amendment as applied to private companies and have the platforms compete. I think we would find most people want to go places where there are not a bunch of racists and sexists and trolls. They want to have a nice time and make art, which I think is the dream.Also, is this as big of a deal as we’re making it out to be? All the issues we’ve talked about have had quite robust public discussions. I will say that the only thing I’m certain about in content moderation is that you make mistakes, 100 percent.Yeah, we haven’t even talked about that part.You always do. It’s humans. Humans are fallible and they will make mistakes. It’s how you correct the mistakes that really matters. We’re at the center of a lot of these stories, as well, with the Hunter Biden laptop stuff that’s now the Twitter files. Twitter decided to remove links to the story to the New York Post. Guess who hosts the New York Post? We do. So there’s some other things where we’re like, “Well, should we take the story down or block the links? What should we do?” We had an internal discussion about that and decided, “This is the New York Post, blah blah blah. It maybe skirts some of the rules we have against nonconsensual hacked material, but it also fits in with these other rules, including being a major-tier journalism organization and public interest.” We made a decision there to not touch it.Wait. So WordPress VIP hosts the New York Post. And when the Post published the Biden laptop story you had to have a meeting about whether to take down New York Post links?There was a discussion. Yeah, absolutely. There is always a discussion and there are reports. People contact us saying, “Take this down,” or, “This is violating your policy.” All of the policies are just a starting point. The interpretation of the policies is really where I think the art and science of it is. We will also make mistakes. We’ve accidentally taken down blogs, either by some script that went wrong, or by a human who clicked the wrong button or made a mistake interpreting our policies. It’s all about how you fix it.I think we’re in a weird period where particularly the right in America is incentivized to say that there’s a huge censorship problem or that they’re being suppressed. Donald Trump would famously play the victim while he was also the leader of the free world, the most powerful person in the United States, the president. That is a shtick that I’m amazed continues to work, but is the problem actually there? Does he actually not have a platform? Is there not a robust discussion around the Hunter Biden laptop? Are there not endless articles, endless testimonies, et cetera? Maybe we just need to say that this is actually working right now, and perhaps we should question the framing that there’s something fundamentally broken or wrong here in the first place. The current system will make mistakes. It’s not perfect, but gets to correctness pretty quickly, usually within a matter of hours or days, not weeks or months. “Elon was in a bubble where he thought this was a bigger deal than it is.”Elon was in a bubble where he thought this was a bigger deal than it is. He bought Twitter and is now slowly recreating all the decisions Twitter has made over the past five years, like taking Ye off the platform or whatever it is. Those are just going to be mistakes that get repeated, and that’s how he’s going to learn, but he’s going to wind up in the same place Twitter largely was prior to him purchasing it.That’s actually the perfect last question. You have done this for the past few years. You bought a social network. You were a very good tech executive when you bought Tumblr. You were very successful with WordPress and all the other companies, so for you to say that this is the most humbling experience of your business career, I think that is very meaningful. You have now done it for three years. What advice do you have for Elon Musk?To keep an open mind, which I actually believe he will do. Whether I agree with him or disagree with him, I believe he’s someone who can update his views when new facts come. We have already seen that happen over the past few weeks on Twitter. I fully expect him to end up where the rest of us are and where Twitter was prior to him. I wish he could have avoided a lot of pain along the way, but do you know the saying, “There are no atheists in foxholes”? I think that there’s no free speech absolutists who run social networks, because you start to realize the nuance of that public square and the responsibilities to the users and society, and the fact that it’s a lot messier.“I hope Twitter doesn’t distract him too much from space, the cars, the solar panels, and everything else.”That’s why you won’t hear me criticize when Facebook or Twitter or anyone else messes up, because I know that we’re going to mess up, too. What I am looking at is how quickly they correct, not whether they are perfect or not, because perfection is not a standard that anyone should be held to. It’s how quickly we course-correct. So that’s what I want to do. I also think he’s working on important things otherwise. I hope Twitter doesn’t distract him too much from space, the cars, the solar panels, and everything else.That’s awesome. Well, Matt, obviously I could talk to you for hours and hours about this stuff. I’m fascinated by the actual experience of running these companies, so thank you for coming on Decoder. We’ll see if we can set a faster record for you to come on next time than we did this last year.Thank you. It was great talking to you.Now to put all this in context I’m joined by Verge deputy editor Alex Heath. He is going to tell me how all of this is connected to Elon Musk and Twitter. Alex Heath, welcome to Decoder.Hey, thanks for having me.So everyone just listened to my conversation with Matt Mullenweg, who is the CEO of Tumblr. The conversation was about what it’s like to buy and operate a social network, and he even said very explicitly, “We’re doing a lot of the things that are on the roadmap for Twitter.” They’re doing payments, they’re doing subscriptions, and they’re going to open-source the algorithm.I wanted to talk to you because you have been reporting deeply on Twitter. I want to just close the loop on some of the things you heard from Matt and some of the things you’re hearing out of Elon Musk’s version of Twitter. Let’s just start with the App Store, which I think is the most complicated and difficult to understand — but also in some cases the easiest to understand — because you just have to do whatever Apple wants. It is a very difficult thing for all of these companies to constantly manage. Tumblr got booted out of the App Store. Matt even said, “Imagine Apple kicking an app from Verizon, one of their biggest partners, out of the App Store. They did it to make an example of Tumblr and their seriousness about content moderation.” Close the loop for me with Twitter. What is going on with Apple and Twitter?I wish we knew more, that’s the thing. Whenever Apple engages in these backroom dealings about the App Store and distribution, even for apps as large as Tumblr or Twitter, we don’t really know exactly what goes down. That is, unless you have someone like Matt who’s willing to talk about it, and even he kind of admitted he’s not really sure exactly what goes down. We know that Elon Musk met with Tim Cook at Apple’s campus, after he had tweeted about how Apple hates free speech in America and how they had paused or reduced spending. They were one of the — if not the — largest of Twitter’s advertisers, which is a strange symbiotic relationship to have with your key source of distribution. Now that beef appears to be squashed, but we don’t really know why. That’s kind of par for the course with Apple. It doesn’t really believe in transparency when it comes to these decisions it’s making about why an app should or shouldn’t stay in the store.Fundamentally, it seems like the content moderation piece for Apple was not as important as it relates to Elon’s Twitter, as the controversy around in-app payments, the 30 percent for Twitter Blue, and all this other stuff that we usually hear about. We have no evidence, and now Elon has even said that Tim Cook told him that Apple was not actually considering pulling Twitter out of the store. Whereas with Tumblr, they actually did. Tumblr was rife with porn and other imagery and content that Apple didn’t like. Tumblr is small and it was owned by Verizon, and the CEO of Verizon is not going to tweet at Apple. Twitter has already skirted the rules for years in various ways. Do you think Elon was using the front of content moderation and free speech to start the conversation about 30 percent? That seems like the heart of it.Well, we’ll potentially know by the time this podcast comes out if they have relaunched Twitter Blue. They’re working on relaunching it as we speak. The reporting that I have seen and heard is that they will try to charge more for Blue in-app on iOS than they will on the web. As far as I know, Apple doesn’t usually like that, so I wonder if that was the thing that Elon and Tim talked about. Only they really know. When they also booted Parler from the Apple store, there were some leaked emails between Parler and Apple that were like, “You need to improve your moderation,” and it was very vague. That’s the thing with Apple. These threats are always vague. So yes, after meeting with Cook, Elon tweeted that there was a misunderstanding about the threat to pull Twitter from the App Store, but maybe Apple was just saying, “We may not approve your relaunch of Blue until you fix X for us.” Actually, I think that’s more likely. As Matt talked about, they usually hold things up in review for arbitrary amounts of time because of other non-related issues.That’s one piece. The similarity between Tumblr and Twitter here is very much dealing with your gatekeepers to distribution, in this case, Apple. Matt was actually much more complimentary toward Google than I expected. My impression has always been that the two app stores are largely the same. He was like, “No, Google is pretty easy to work with.” At some point, the two big gatekeepers exist and you have to manage them. We have seen that Tumblr has had its challenges, and we have seen Elon run into those challenges. Then there’s what they’re actually trying to build towards. It’s remarkable how similar they are. They want to get away from a pure dependence on advertising, and they want to launch paid consumer products like Twitter Blue or tipping on Tumblr. Tumblr actually has fake verified badges, which is one of the funniest social media products in years.How is that going for Twitter? This stutter-step towards Twitter Blue exists, but Tumblr is a much smaller company and network, so they can move the numbers on the revenue much more easily. Twitter is bigger — it’s not the biggest, but it is bigger — and like 85 percent of their revenue is advertising. It has to make money. Elon has to pay debt, whereas Matt told us Tumblr is still losing $20 million a year. How hard is it going to be for Twitter to pull this off?It’s going to be incredibly hard. They have to really identify what people like us will pay the most for. Yes, Twitter was experimenting with subscriptions before Musk came in, but it was languishing as a product. I was a subscriber, and I didn’t feel like I was getting much value out of it. I would be curious to know if Tumblr is the same way. Does it have this weird effect where a very small percent of the users produce most of the content? Those are the people that you could extract the most money from, because they’re doing that for a reason. It’s because they’re getting something in return. Where can you chart that value? How can you make that into something they will pay for? “Musk really needs that to pay the bills and to pay back the interest on the largest leveraged buyout that any individual has ever done.”You’re right that it’s harder for Twitter because they have this existing multi-billion advertising business. Musk really needs that to pay the bills and to pay back the interest on the largest leveraged buyout that any individual has ever done. I’m not sure that they can do it. From what we have seen, guess what? Advertisers want a certain level of brand safety and moderation that so far Twitter under Musk has not been able to provide. It sounds like Matt has realized for Tumblr it’s going to make more sense to not be reliant on advertising. I don’t know if it’s for that reason, but there’s also this element where the scaled ad play on social is kind of over. Even Meta is working on paid products now for its apps for Instagram and Facebook. I don’t really see any of the more upstart social companies focusing on a scaled ad play. I think we’re past that era.What about the payment side of it? They’re all talking about payments. They want to be able to make you send money to other people on the network. It seems like, “What if a bunch of people are sending money around and we took cents out of every transaction?” I get why you’d be interested in that, but it is also the most boring product for a social network. Also, I don’t know that I want to be sending money on 50 different platforms.It depends. If you have a thriving creator system where creators are posting more content and asking for payment, maybe you do want to have money in the system. Elon is obsessed with recreating his original idea for X.com, which predated PayPal. I reported on a meeting recently where he told employees that PayPal was just phase one of what he actually wanted to do. He has every aspiration and intention to complete this Project X, as he calls it on Twitter, which is to turn it into a bank. No one has done that successfully.If I were to stack rank all these insane challenges he has ahead of him, that one just seems so lofty and hard. You have to have all these other things figured out first before you get there. So I don’t know if or when that will ever happen. For Twitter to be the $100-plus-billion-market-cap exit that it needs to be, for him to make good on the investors who put hundreds of millions of dollars into this buyout, he needs to have more than just Twitter’s current ads business. So it’s payments. That’s all there is.That was going to be a follow-up question. Is there anything else? Is it, “Okay, I saw a good tweet, I’m going to kick a couple bucks towards the person who wrote the tweet”? That’s the baseline of it, but I’m not sure why I would do that.I don’t know, put it on the blockchain and it’s gas fees all the way down. I mean, no, that’s it. You have ads or you have some version of payment / subscription. Musk has been very clear that Twitter needs to be at least 50 percent subscription or it won’t survive what he thinks will be a very painful recession that will affect ad spend next year. He has been very clear about this in multiple internal meetings lately. He probably has pretty good economic data in front of him and people telling him what to expect, so I have to give some level of credibility to that fear. “I do think the future of social is increasingly subscription, whether it’s Twitter, Tumblr, or Meta that figures it out.”I do think the future of social is increasingly subscription, whether it’s Twitter, Tumblr, or Meta that figures it out. We’re seeing it across the creator economy. Everyone wants that direct relationship and to not be disintermediated by an advertiser. At the same time though, Twitter is always going to have Apple as that intermediary until something happens there.Yeah, the Musk / Apple relationship right now appears to have played out as he tweeted angry things about Apple reducing its ad spend, in-app purchases, and free speech. He then went to the Apple campus and they had some sort of conversation. Musk told employees in a public Twitter space that Apple’s spend was all the way back up, and he has stopped complaining about the 30 percent fee. He is just going to spend Apple’s ad dollars right back to Apple. That’s pretty funny. It’s hilarious that the money is just going in a circle.It’s good to be the platform, right? Apple’s bread and butter is extracting money out of the apps that sit on top of its phone. I’m curious, when Matt was talking about this, does he feel like Apple deserves this money? Did you get that indication from him?I get the indication that every CEO knows that there’s a line, and that they are willing to walk right up to that line. The line for Matt is clearly farther than most other CEOs that I talk to. They are not willing to go over that line. Matt is willing to say, “Apple has a lot of power, they hold us up in reviews, and we get it, because there are nipples on Tumblr.” They recreated a system to allow nudity on Tumblr by putting toggles on the web. They’re like, “This complies with Apple.” He’s willing to talk about that. I don’t think he’s willing to go one step further because Apple can destroy his business. Over and over again on this show, we discover the line of what CEOs are going to say about Apple. I think that’s just utterly fascinating. I think the thing with Elon that is fascinating is that line does not exist for him.He’ll just say whatever he wants. To some extent, that has been very illuminating throughout this entire process.I think what Elon has shown is that the conversation is shifting from, “Apple’s control is a business issue for everyone,” to, “It’s actually a speech issue.” We’re seeing Tim Sweeney and other CEOs kind of pile on this. I think this is the next phase. If Ron DeSantis and Elon are saying that you’re threatening free speech, you may have a problem that you need to combat, even if it’s just a PR one. I’m not sure Apple is equipped to engage with that level of attacking.I think they’re ready for it. I think what they’re going to show is, “Look at these apps. They’re full of bad things you don’t want your kids to see. We sit in the middle and make sure that your kids don’t see that stuff. If you want to see that stuff, go use our web browser.” I think that has fundamentally been their answer for a long time.iPhone, the phone for good parents everywhere.It’s their brand. This leads us right into content moderation, which Matt and I talked about at length. Elon is trying to restart from first principles at Twitter, for better or worse. We can talk about the Twitter files at some length if you want to. Matt’s point was that he is going to end up right back at the beginning. He said this several times in this conversation. I found this striking, because Matt is a good CEO — he is a long-tenured, extremely well-regarded, extremely effective CEO in Silicon Valley who makes a product that millions of customers use. He was like, “Buying Tumblr has been the most humbling experience in my career.” A huge part of it is the content moderation piece. His perspective is that he’s a very libertarian-leaning person when it comes to what people are allowed to say, even if on some other issues he’s more to the left. He was like “On speech, I’m libertarian. To run Tumblr, I can’t do that. We have to shut a bunch of stuff down.” He brought up “lawful but awful” and all these sorts of tropes that we hear from the trust and safety community. He was like, “Elon is going to rapidly find out that he’s going to start somewhere, and he is going to end up right back where Twitter began.” I think the Twitter files are somewhat of an exercise in this. Elon is out there saying it’s freedom of speech, not freedom of reach. Then the Twitter files are Twitter implementing exactly that idea in various ways to a first approximation. Do you think Elon is going to end up right back where he started, or do you think he’s going to end up someplace radically different?“The mantra inside Twitter is that you could essentially say the most hateful thing, and unless it’s illegal, it’s going to be on the site.”He’s not there right now. The mantra inside Twitter is that you could essentially say the most hateful thing, and unless it’s illegal, it’s going to be on the site. Our job now is to not amplify it, to not suggest it in the timeline, and to basically corner that speech off to the follower graph of that account. As Matt told you, that’s a very nice libertarian view of speech and how it should work, but it’s not how an ad-supported platform can function at scale, at least from what I have seen. Elon, his head of trust and safety, and the people implementing this stuff, they’re not there yet. I know this for a fact. Will it take the advertiser business of Twitter completely crumbling for them to realize that? I don’t know. Right now their baseline for success is, “We are not amplifying hateful, racist, misogynistic tweets and we’re not putting them next to ads.” That hasn’t happened yet, but that’s what they want. They will think they’re adhering to their “freedom of speech, but not freedom of reach” principle, and no platform has shown that that’s enough.Also, who wants to be on a platform with bad people? That’s the weird part to me. It’s funny. To some extent, with overt racism, overt sexism, and overt transphobia — and Matt brought up the pro-ana community, which says, “anorexia is good” — on the whole, people are like, “Yeah, that stuff is bad.” Then there’s a lot in the gray area. Even the stuff that people agree is bad, people don’t want to be on platforms where that stuff is prolific. So if you need to grow the user base and have payments, don’t you need to do more than wall it off? Don’t you need to just make it go away?You would think so, especially if your goal is to be the town square. It turns out that if you actually think about the digital representation of what an actual town square would look like, it’s not a good place to be. It’s everyone from the town in one square yelling at each other, which is what Twitter has already been. They have earnestly been trying to get rid of the worst voices in a way that seems measured based on the own internal correspondence that Elon has had his friends trying to disseminate with the Twitter files. Maybe that is what’s going to have to be discovered — that this whole town square concept just doesn’t work, because humans don’t actually want that.You said the Twitter files have been disseminated by Elon. It’s unclear how they’re being generated or vetted. There are a lot of question marks there. What they basically show, from what has been publicly revealed, is well-meaning people earnestly debating difficult decisions and arriving at some conclusions. Maybe you disagree with that entirely. Maybe you think they’re not well-meaning. Even if you think it’s a shadowy liberal conspiracy, you can’t really disagree with, “Yeah, they’re talking about hard decisions and reaching a conclusion, while also talking about how to justify their conclusion.”Then there’s this concept of shadow banning and limiting your reach, and what you just described is exactly that. “We’re going to detect the content of your tweets and make sure we don’t show them to anyone.” Maybe you will know, maybe they’ll be more transparent about it, but they’re going to limit you because they don’t like how racist you are. That is a very qualitative, very difficult kind of judgment. I don’t think that you can automate it. Do you have any sense of how they will actually implement that?No, and they don’t know. They hope to automate the worst of the worst, but you’re right, there’s so much nuance and tone. There’s no platform that is doing this automated de-amplification of nuanced, potentially sarcastic but hateful speech at scale. I think it’s deeply ironic that as he’s tweeting “freedom of speech, not freedom of reach,” he’s having these cherry-picked files being dumped that show them doing exactly that.I guess with the new Twitter 2.0, the hardcore Twitter, he’s trying to make a point that it will continue, but it won’t be politically motivated. I guess the insinuation is that the people before were doing this because of their political leanings. I think there is some cringe stuff that former Twitter execs did and tweeted about that showed leanings that they shouldn’t have exposed in that way. But it’s the same thing. It’s like he’s saying, “I’m going to keep doing this, but you like me, so therefore it’s going to be okay.”Once you’re the head of a social network, you’re the benevolent dictator, whether you want to be or not. I think Elon’s like, “Screw it, I’ll just be the dictator,” in a way that Jack Dorsey kind of didn’t want to be, for example.He tried not to be. Almost for the worse. He tried too hard to not be involved, and we’re seeing that now. It’s like, “Where was Jack?” That’s a whole other tangent though.Is there anything else you picked up from the conversation with Matt that you think is applicable to Twitter?“As a culture, we haven’t even really come to reckon with the ramifications of the power of where you sit in the stack and the content decisions you make.”The Twitter files have centered on the Hunter Biden laptop story, and I thought it was fascinating to hear Matt talk about that. I didn’t know they hosted the New York Post, but that they even debated potentially pulling that... As a culture, we haven’t even really come to reckon with the ramifications of the power of where you sit in the stack and the content decisions you make. What a company like WordPress could do is arguably even more of a powerful thing than taking down a tweet of a story.It’s important that we do more to peel back how these companies actually deliberate, and either almost or don’t do something like that. That’s the closest I can come to agreeing with how the Twitter files are being handled. I agree with the core nut of that idea, but the way it’s being done is not great. But I’m glad Matt talked about that because it shows these discussions happen that we don’t really know about, and that could really impact speech on the internet.If you host a major American newspaper, you should probably treat these platform companies differently than something else. Those decisions are not transparent to us. I don’t even know if they’re transparent to the New York Post or whatever. I thought that was really remarkably telling and brave of Matt, as the CMS provider of the New York Post, to say, “We have the ability to take their links down.” We actually talked about it in the case of the story.I do think it is important for everyone to remember now that we’re deep into it. You can just get Hunter Biden’s laptop if you want it. Apple sells it in the store. There was a time when no one understood the provenance of the laptop, and no one understood what was on it. Mostly it was a bunch of non-consensual nudes being shared, and people thought that it was a Russian operation. The over-heatedness of that moment probably led to that conversation. It is also remarkable for the technical capability of that conversation to even exist.And that we are still talking about it. I guess it’s because it’s such an uncomfortable thing that could have happened and did happen on Twitter. When something is super political, we realize it in the heat of the moment like, “Whoa. We actually have these platforms that sit at various layers of the stack that have tremendous power to literally just wipe that off of the internet. What happens if they actually do?”It’s fascinating to me that as a CEO of WordPress, he wasn’t like, “That was the most humbling moment in my career.” Instead he was like, “Being the CEO of Tumblr and owning Tumblr is the most humbling part of my career.” Because then you have a mass of users. You have millions of unpredictable users doing whatever they want, and you have to somehow control them.You should start mailing a “Welcome to Hell” PDF to every tech CEO that runs a social site.Matt didn’t like that headline, but he told me he liked the piece. It was very validating. All right, this has been a fascinating episode. I am curious to see how fast Elon comes back around to the baseline of operating a social network. Matt’s a smart guy. Zuck, for all of his faults, is a very smart person and he has arrived at a place that looks a lot like the place Twitter was at. If the constraints are such that all these smart people sort of arrive at the same spot, I’m curious to see if Elon arrives at that same general position in the end.If he doesn’t, I don’t see how Twitter has an advertising business. Those are going to be the two sides, I think.For sure. All right, Alex, thank you so much for coming on Decoder. We’ll talk to you soon.A podcast from The Verge about big ideas and other problems.SUBSCRIBE NOW!

Transportation & Robotics

An expert in aeronautics discusses how recent developments in fuels, engines, materials, and computer controls are leading to an era of airplanes unlike any that have gone before.

Our guest on this episode of Stanford Engineering’s The Future of Everything podcast, Ilan Kroo, is an expert in aircraft design. But when Kroo talks of aircraft, he means a new generation of flying vehicles that could transform our very concept of transportation—like personal flying cars that take off and land vertically or commercial airliners fueled by clean-burning hydrogen. Kroo says the rapid changes he’s seeing in the industry could lead to safer, less expensive, more efficient—and quieter—air travel. Listen in as he shares more with host Russ Altman on this episode of The Future of Everything podcast. Listen and subscribe here Related: Ilan Kroo, professor of aeronautics and astronautics

This week’s guest is Jeremiah Lee, who was previously a manager at Stripe and product manager at Spotify. This conversation focuses on org structure, and specifically Jeremiah’s experience with the popular squad model from Spotify. Jeremiah provides the backstory on where the model came from, what parts of the model were a challenge, and advice for leaders either already adopting the model or considering doing so. ‍Discussion points:(1:40) What the Spotify model is(4:39) Jeremiah’s impression of the Spotify model as he joined the company(7:29) Spotify’s progress in adopting the model as Jeremiah joined(9:55) Challenges with matrix management(12:02) The role of engineering managers (14:40) What the model was designed to solve (15:54) Good autonomy versus toxic autonomy (18:51) How Agile coaches were used at Spotify (21:39) Advice for teams who are struggling to implement the Spotify model(24:50) Advice for leaders who are starting to think about org design(27:30) How Stripe approached org structure (30:26) How org structure affects a platform team’s work (33:32) Tracking engineering org structures (36:02) Why the squad model became so popular(39:37) What the original authors may have felt about the popularity of the model‍Mentions and links: ‍If you enjoy this discussion, check out more episodes of the podcast. You can follow on iTunes, Spotify, or anywhere where you listen to podcasts.Abi: Jeremiah, thanks so much for sitting down with me today and taking the time to come on the show. Really excited to chat.Jeremiah: Thank you so much for having me.What the Spotify model isAbi: Well, I'm really excited for today's discussion because we're talking about org design. Org design is something that has such a big impact on developer experience and developer productivity, and it's really something that we get asked a lot about from leaders and that we think more leaders should be thinking about in terms of ways of optimizing the efficiency and productivity of their organizations.I know you have so much insight and personal experience with this, particularly with the Spotify Squad model. And so I'd love to just dive into that. To start off with, for those who are not familiar with it, can you give a quick primer on what the Spotify Squad model is?Jeremiah: Sure. So in 2012, Spotify published this blog post in a video about its way of working, and that became known as the Spotify model, even though Spotify didn't call it that. It's basically two ideas. The first is stream aligned teams, and the second is matrix management. And then they threw in some weird word choices just for fun. So Spotify had teams that it called squads, and a group of teams were organized into a department that it called a tribe. And then a business unit was called a mission. And each team was intended to be this autonomous mini startup. So it had a product manager acting as a mini CEO for an entire feature area.These teams had all of the designers and software engineers with a range of specializations that they needed to deliver a feature. The idea is that they should have no dependencies or not have to rely on any other team and just focus on shipping that feature, go be successful.So that's not really a new idea. It's what we would call a full stack team, multidisciplinary teams or agile feature teams back in the 2000 ops, or we might call it a stream aligned team today. But the other major element of the Spotify model was matrix management, and this is a much less common one. So instead of engineering managers, there were chapter leads. So these chapter leads would manage just a specific type of software engineer across a department. For example, you might have all of the software engineers who are working on backend engineers across all the teams within a department. They'd have one manager. And then all of the Android mobile engineers in the department would have a different manager. And so even though that API engineer and that Android mobile engineer worked on the same team, they would have two different managers. The intent for this was to allow engineers to basically move between teams within a department to best meet the business requirements without them having to change managers. It was well-intentioned, but I think had some drawbacks.Jeremiah’s impression of the Spotify model as he joined the companyAbi: For sure. And we'll get into some of those challenges and drawbacks later. But I remember when that article first came up from Spotify, it did seem very new to me. I remember, like you said, the concept of having these highly autonomous and flexible feature teams, that was very intuitive and really inspiring and positive. I remember the piece about decoupling management structure from the functional work being done was more interesting, and I think snowballed into a lot of different approaches in the industry today, which we'll talk more about.When you first learned about the Spotify Squad model, when exactly was that and what was your personal impression of it when you heard about it?Jeremiah: Sure. I first learned about the Spotify model when I worked at Fitbit a few years earlier. At that time, Fitbit had been growing itself. I joined when the company was around 150 people and in just a few years, it had grown to around 700 people. And then it had a bad situation. For a variety of reasons that I can't get into, it unexpectedly had to launch three hardware products simultaneously, and that nearly killed the company in terms of just burning out the entire company in order to do those three simultaneous launches.So the leaders at Fitbit knew that they had to do something differently, and the Spotify model was one of those ideas that they discussed. So that was the first thing. And I think the idea that really resonated with Fitbit and its leaders at the time was really its first exposure to the stream aligned feature teams because our teams were organized primarily around discipline at that time.Abi: That makes sense. So around that same time, I know you started looking into joining Spotify and then interviewed at Spotify. Maybe share a little bit more about that experience. I'm curious, were you partly attracted to Spotify because of the Squad model that had been published? I remember you mentioned you had learned some things about it throughout the interview process itself, so I'd love to learn more about that.Jeremiah: So I actually left Fitbit because my husband got a job at Spotify and we moved to Sweden from San Francisco. So that was the original movement there. And then they were going for a two for one deal. I also interviewed at Spotify as well after I moved to Stockholm. But you're right, I actually was interested in working at Spotify myself because I had heard good things and it was a great product. It had a great reputation and I tend to be attracted to companies that make products that I think are cool.So in the interview process, it was a part of the interview where I was going to get to talk with other engineers on the teams that I'd be working with. And I was like, "Oh, I'm really excited. I can't wait to talk to people to hear more about how the Spotify model actually works in some detail." And the recruiter, she was like, "About that, I think you'll find that we work quite differently than what you might have read on our blog." And that only intrigued me more. So I did end up asking engineers in the interview process and they all confirmed that, "Yeah, we use those weird words more than the ideas," and that collaboration across the company was still one of the most difficult challenges that they faced as engineers inside of the company. Yeah.Spotify’s progress in adopting the model as Jeremiah joinedAbi: That's really interesting. As far as what they were actually doing, what was it then? In what ways was it aligned with the Spotify Squad model and what ways did you find it was different?Jeremiah: Oh wow. Okay. So at that time, Spotify had more than doubled the size of the company within a year. And so they were going through this hyper-growth phase, and there was just a lot of everything going on. One of my favorite questions to ask when interviewing at a company is, "Tell me how something goes from idea to value delivery to a customer." And every person I asked that question to, gave me a completely different answer. And I think that's just really indicative of the strong autonomy culture inside of Spotify. So I'd say as far as things in the model, the thing that was most reflected in reality was just this extraordinary amount of autonomy that each team had. So that was true. And then of course, the unusual word choices they did call things squads and tribes. I still don't understand that one, but that was true.They had started to realize that the matrix management wasn't working, and so they were starting to move to having chapter leads... They still kept the name, really be focused on a single team in managing all of those engineers within the team. It would actually be quite a challenge internally to get those renamed as engineering managers. That had started to shift when I was there. And then there was still so much, I guess, bottom up movement. Everything had to basically be done from the individual team level and build up. And teams got these very high level directions from above with the system that they called the bets system. So these are basically, here are the top priorities for the company. Some things would get some extra focus and staffing and whatnot, but there wasn't much happening there in, there's a growing middle section of the company to really drive alignment inside of the company.Challenges with matrix managementAbi: That sounds like almost the middle layer of the management structure was too loose, not enough alignment from top to bottom. So it's interesting you mentioned that it seemed to be there was growing recognition across the company that there were some challenges with the Squad model. You've also, of course, written this article about your personal observations. You touched on that already, but you mentioned that matrix management solved the wrong problem and there were challenges with it. So can you elaborate on that a bit further?Jeremiah: So matrix management makes this assumption that you're going to have these self-organizing teams and that they're going to be constantly reorganizing. Well, I just don't think that's true for most companies in the way that they work. And that certainly was not true at Spotify, at least at the time that I joined it. You had long-lived teams that were basically consistently delivering with fairly long roadmaps within a particular area. So there was some movement happening within teams, but it's more about shedding responsibilities or being able to focus on things as you got more staff than it was about the whole team reorganizing. So that's just basically this one assumption.The other aspect of this is what you just mentioned about that middle. It's like, well, what should the role of managers be inside of this organization? And I think there was some general skepticism of the value of management in general, and that may have been okay when the company was smaller, but as you get bigger, I think management is important. Maybe it's just me now working primarily as an engineering manager. But we need, those chapter leads or engineering managers, we need them to be more than career coaches and we need them to be more than servant leaders with no delivery responsibilities. I think that's actually dismissive of the value of managers. Managers, I think, should be responsible for the performance of a team, not just the individuals on the team and not just the careers of those individuals on the team. Also, most of the problems that we tend to see within engineering teams are rarely technology problems. Those are people problems. And it's really hard to solve people problems if you're not actively involved on a day-to-day basis with all of the members on the team.So I think there might be some situation in which matrix management can work, but it certainly was not working for Spotify when I joined, and I just haven't really seen that situation in my own career arise where it would work well for an organization.The role of the engineering managerAbi: Yeah, really interesting points. I want to follow up on this last point first, which is around the fact that what's the actual role of a manager and can they be effective if they're not actually integrated into the work a team is doing. In your opinion, is it an anti-pattern to have engineering managers who are only people managers and not involved in work? Because that something I think I've seen adopted in organizations even that aren't trying to be prescribed to the Squad model. It seems to be a trend that was maybe born out of the Squad model that I've seen across the industry. So what are your thoughts on that?Jeremiah: I don't think there's one right answer for what a manager should or should not do. But I would go back to why do we have managers? What is the value that they are bringing to a particular team and to the organization? My perspective is that a manager should be the one that is accountable and therefore, responsible for the delivery of the team. And in order to have quality delivery on a team, you need to have someone who understands the needs to accomplish the work, who understands the competencies of the people on that team and where those gaps are and where the growth opportunities are, and where you might need to change staffing in order to fill in those gaps. And I just think that would be very difficult to do if you don't have some level of understanding of the details of the work that has to get done and the career paths of the people who are going to be doing that work.What the model was designed to solve Abi: I'd agree with you on that, and it's something I think you see that disconnect arise in the same way that you observed at Spotify, in lots of organizations that do decouple those responsibilities.Earlier you talked about how one of the things that is inherent to the Spotify Squad model is this concept that teams can constantly change and reform and restructure. And in fact, in a lot of organizations I've seen that implement the Spotify Squad model, that becomes a part of the way they work. For example, I know of one organization that every quarter, reforms their squads based upon the projects for that quarter. I know your experiences that teams are often long lived, and that's been my experience as well. But at Spotify, was this designed to support the way that teams are already working or did teams start reforming more often because of the Squad model? I guess I'm asking, was this designed to solve for the way that organization was already working at Spotify or was this a little disconnected in your view?Jeremiah: This one is difficult for me to answer because I wasn't there at the time when the document was written. But something that I did notice at Spotify is that, up until recently, Spotify solved basically every problem that it had by hiring more people. It's one of these companies that is in this luxurious position where it can continually hire more people to solve particular problems.So the reorganization that I observed when I was at Spotify tended to be like the company is growing and with more people, that allows us to have teams that are increasingly more focused on particular areas. And so that's why the shuffling of people on teams was happening, not necessarily because there was someone being, "Reshuffle yourselves," every quarter or something like that.Good autonomy versus toxic autonomy Abi: Right. There's just a lot of growth and natural reformation as a result. That makes sense. I want to move on to another point you bring up in your article, which is this fixation on team autonomy. I can guess as to where that leads, but would love to hear more from you on what the problem is with that?Jeremiah: Okay, so before I ever criticize autonomy, I want to say autonomy is good. Autonomy is a requirement for individual happiness in general. You want to be in control of your life, in control of your destinies. And I think that there are a lot of companies that suffer from too little autonomy on the individual contributor and the team level. So autonomy in the Spotify model, I think is very much influenced by Swedish society. It's an aggressively egalitarian society, and I think hierarchies in general are treated with suspicion. And again, given all the inequality and inequity in the world, it's probably a really great starting point. But it is a spectrum of things and autonomy is something that needs to be balanced out. So in general, my feeling is you have autonomy, which is, am I able to use the best of my ability is to solve this problem? And am I respected and consulted in the making of a decision that affects me? That's good autonomy.But you can also have toxic autonomy on this other end of the spectrum, which is like, I don't take direction and my needs are more important than the team's needs, or my team's needs are more important than the organization needs, and I don't want to collaborate with you and no one's making me. So that's the other spectrum of that. Well, this is a business. It's not a commune. We do have to have some structure and we do need to value collaboration in our collective progress over individual or team progress. And I think trying to find that right balance is a difficult thing to do and it's one of the biggest challenges that I witnessed when I worked at Spotify.Abi: That's really funny. I formerly worked at GitHub, which didn't have Swedish roots, but funny enough, the history of GitHub, it also went through a multi-year period during its rapid growth where the development organization was against the idea of having managers. There's been a lot written about this era, but it took many years thereafter to fix that problem and bring in a management layer. To be honest, it's something the organization is still adapting to and struggling with today. So it's really interesting to hear about the backstory at Spotify.Jeremiah: Yeah. I also witnessed this within Stripe. I think autonomy, again, it's a really good thing, but it's also possible to have too much of a good thing. And especially as you grow, you do have to start thinking about how you balance autonomy with alignment and accountability. And both of those things get more difficult as an organization scales.How Agile coaches were used at SpotifyAbi: Absolutely. One other thing I want to go to from your article, you talked about teams facing challenges with collaborating, and you also talked about the role of Agile coaches at Spotify. So I'd love for you to elaborate more on how did the matrix structure affect team's ability to collaborate successfully? And for people who aren't as familiar with Agile coaches, how did Spotify use Agile coaches to try to help bridge that gap or alleviate some of the challenges?Jeremiah: Sure. Again, before I start criticizing Agile or Agile coaches, or critiquing it anyway, I want to say in general, I believe that Agile is the best guiding principle for creating software that we have. It's a set of principles that I believe in and I think help make great organizations, but there's a lot to learn there. And I believe that collaboration is specifically a skill in structuring work and problem solving within software. Those are also skills. And they're not necessarily things that you get in a university setting. They're primarily things that you infer from working with people or being in a particular environment.So Spotify, like a lot of places I've been in, had a bunch of people who didn't have a basic understanding of those Agile practices, and yet they were so autonomous that they also had the responsibility of figuring out. So Spotify had a whole bunch of Agile coaches who would basically float between teams to swoop in and see how the team was doing and try to identify some problems and suggest some things that you might try to do differently to basically find better ways of working. Not bad in concept, but I think this is one of those things where you need a longer engagement to effectively do that.In this case, the Agile coaches didn't have any... They weren't effectively, were not responsible for anything because they were just floating across the organization. So they weren't dedicated to making sure that the team's delivery practices were actually creating positive results for the team, for the company.So you need to educate people and you need to have some system in place for teaching people and giving them the knowledge and the vocabulary to understand how these practices can be applied and how to use them and how to evaluate if they're being effective or not. But what you don't want to do is just keep throwing ideas at teams where they're like every few weeks or month or whatever, they're just iterating, "Well, that idea didn't work, so let me try this other Agile idea." That that just doesn't work.Advice for teams who are struggling to implement the Spotify modelAbi: Yeah. Reinventing the wheel and also probably trying to train developers on Agile process things that they're not maybe that interested in or passionate about, sounds like it could be a challenge as well.We've covered so many aspects of how the Squad model runs into challenges and also how Spotify itself was facing challenges with the Squad model. Some of the takeaways so far, at least for me, have been the importance of having managers be responsible for delivery, not just people. And also tempering the balance between autonomy and standardization across an organization in terms of teams. So many organizations out there have gone ahead and adopted the Squad model based on the blueprint that Spotify published, and many other leaders are constantly asking whether they should. So I guess for these organizations out there, especially those that have already implemented it and it's maybe not going great, what would be your advice on what they should do?Jeremiah: So the premise is if a company has adopted the Spotify Squad model, it's not going well. This is not an unusual thing, I think. I've spoken to several companies that say like, "Oh yeah, we adopted the Spotify model." And I always ask, "Well, how's it working out for you? And tell me about how you did it." And the reality is there's not much there there inside of the Spotify model. There are just not enough details to effectively replicate it to many organizations. So every implementation I've heard of has been slightly different, but I do tend to hear people having the same problems when they try to do it.So if you're a company and you've adopted the Spotify model and it's not going well, first of all, I want to say good job on your willingness to try something new and for the self-awareness to evaluate that things are not going as well as you'd like because there are so many organizations that can't do either of those things well.But I'm going to ask you to compile a list of the problems that you're observing and to interview people throughout the organization to try to deeply understand those problems and the causes before you make any more organization changes. And I am going to recommend that you read Team Topologies by Matthew Skelton and Manuel Pais. And if you're really bought into the idea of matrix management, I'm going to challenge you and I'm probably going to recommend some other books on management to think about that one a bit more.But I want you to try to figure out the types of work that need to be done in your company and then try to find the right team structures to get that work done. But before you implement any more changes, you need to be talking to your best performing individual contributors about that plan and get their feedback on that. Because what I've seen a lot of companies, there are these invisible support structures and processes within organizations that work really well, and it's usually the things that are not being prescribed to people, but they're basically the channels in which work actually gets done well. And as a manager, you want to find those. You want to make those visible and hopefully reinforce them with an organization structure and some new practices that you might introduce and just do a bit more research before you start making any more organization changes.Advice for leaders who are starting to think about org designAbi: Well, I think that's great advice for leaders who are already living within the Squad model in their organizations. What about leaders who are just beginning to ask that question of what's the best way to organize and design our org? For them, would your answer be to go read Team Topologies, and then more thoughts on how to approach that matrix management structure based on your own experiences?Jeremiah: Sure. Yeah. I think Teams Topologies is a fantastic book because it basically outlines four different types of teams that you're going to see in any company that is producing software. You can basically use any organization structure with fewer than 150 people, you have this Dunbar's number, as a company starts to grow. Then you have to start being a bit more methodical about how you start to do alignment and accountability and autonomy inside of an organization. But anything below 150, it's probably going to work.So it is a scale and a growth problem. But Team Topology basically gives you a little bit of a blueprint for the types of teams for the different types of work that happen inside of an organization and the number of each type of team that you're going to need changes as a business grows. If you're a really small company, you probably don't need a whole bunch of enabling and platform style teams, but as your company grows, you need to start building up human infrastructure, and basically these are teams that are supporting other teams inside of a company.And so it's good to understand what your growth trajectory might look like so you can start to plan for the types of skills and the type of people infrastructure that you're going to need to put in place in the future. And the other thing is, again, if you want to learn more about Scrum, the other book that I constantly recommend is Essential Scrum by Kenneth Rubin. He actually did all of the training on a quarterly basis at Fitbit for teaching people how to do Scrum. He's a really great educator. In his book, lays out the ideas in ways that I think are very, they're focused on delivery in a way that engineers can really buy into.Abi: Thanks to that book recommendation. Was that the underpinning of the Agile philosophy at Spotify as well or was there a different set of books and people that Spotify was more designing around?Jeremiah: Spotify had more Agile coaches than any company I have ever witnessed, and they were all rightfully really educated and up on the latest research and thinking as well. So there was a lot of stuff going on inside of Spotify, including several people who actually run consultancies and have written books and stuff on those now.How Stripe approached org structureAbi: I'd love to switch topics and talk about your experience at Stripe, which I know you recently left there. Stripe is an organization that a lot of people look up to in terms of their, at least from the outside in, their development acumen and thought leadership in terms of just where the industry is headed. So naturally I think a lot of listeners will be interested in knowing how does Stripe structure an organized team? So we would love to hear what your observation was on that while working there.Jeremiah: I'm going to caveat this way. Stripe has an extraordinarily restrictive policy for what its employees are allowed to say publicly. So even though I don't work there anymore, I still want to respect them in that regard. I will say that Stripe does have a more traditional organization structure. You'll see a lot of the four types of teams subscribed in Team Topologies inside of Stripe, but it also went through a hyper-growth stage. During the pandemic, it more than doubled in size and even though it had this really great strong culture and some really good practices internally that really supported remote and asynchronous work even before they needed to do it because of the pandemic, scale will always present new challenges. And so the biggest thing that Stripe was figuring out was still how do you do alignment now that you're more than 6,000 people? How do you do effective decision making? These are common problems inside of any organization and it's really just trying to understand, again, how do you balance autonomy with alignment and accountability?Abi: One of the benefits of the Squad model, one of the challenges as well, is that teams can change very quickly without having to change management structures. So I'm curious, under Stripe, I imagine there are many long lived teams, but in cases where teams did need to change, would that typically involve a manager change as well or would it follow the Squad model where the project teams would change without changing the underlying management structure?Jeremiah: So I have to say, I was reorged three times within a year working at Stripe and every one was justifiable. They were shifting priorities and I generally agreed with the need to do so. So it's really hard for me to say what the norm was in that situation. I would say Stripe understood the value of team cohesion and did as much as possible to try to keep effective teams continuing to work together even if the scope of the work changed. And managers play a critical role in that team cohesion. So again, as much as possible, you want to keep managers with the team and I thought Stripe did a very good job at that.How org structure affects a platform team’s work Abi: That's good to know. At Stripe, I know you ran a platform team, and of course most of our listeners of this podcast are leaders of platform teams or DevX teams or developer productivity teams. In your experience, having worked at all these different types of organizations and then being a platform leader, how does the way that an organization is organized in terms of its team structures either benefit or create challenges for the work you do as a platform leader?Jeremiah: This is one of those things where, when you're on a platform team, I think the biggest challenge that every company I've worked at has basically been getting priority in all of basically your internal customer teams backlogs. So as a platform, you're providing some particular capability that a bunch of other teams depend on, and you do as much as possible to not create work for those teams, but you're going to be in some situation where you're going to need to do a migration or you're going to need to do an upgrade and you're going to have to push work back to those internal customer teams.But in those situations, you almost never have authority to prioritize that work for those customer teams. And so you have to find a way of being to influence that prioritization. And this is where I think product management techniques are really helpful. You need to basically be able to explain the business value of doing the work that you're asking and you're trying to connect a company goal to the work that you're asking teams to do. And in some cases, you might get some top-down help to bump up the priority of things in the backlog, but you really need to understand what are the key business value things that we're driving for, how are we measuring the effectiveness of this. Yeah, so I think in general that's like the biggest challenge just managing any sort of platform team.‍Abi: And I imagine that challenge looks a little bit different if your boss is the same boss of those feature teams that you're trying to convince. So I'm curious, that probably makes the job easier, but in cases where you're working with teams that are in completely different chain of commands within the organization, that must be a lot harder. Right?Jeremiah: It absolutely is. But if there is value in doing the work, you should be able to communicate the value of that and then you should be able to compare the value of that work versus the other work that also needs to get done. If you go looking for work, there's always work to be done. But we have to be able to, I think is one of the big things about management, is you have to be able to connect the work that you're doing to the business value that you are creating. And it might be related to more revenue, it might be related to reducing costs, it might be related to reducing risk for the business in some way. So I think it's extremely important that managers are able to connect work to the business value and then that should give you at least the basic tooling to understand why this work should be done versus other work that's in a backlog. And then a good organization is going to have ways for you to be able to escalate if it's necessary.Tracking engineering org structures Abi: Speaking of connecting work and connecting the dots between teams and initiatives, one of the challenges I see brought up a lot is actually tracking org structures, particularly when they are somewhat matrix and decoupled from management hierarchy. I think almost all companies have an HR system that tracks who reports to whom, but seems very few companies have some sort of up-to-date system that's capturing what the teams are and what they're maybe colloquially called within the organization. So I'm curious, at Spotify and at Stripe, did you guys have any novel solutions to the tracking problem around teams?Jeremiah: Yes, actually. This is a really great question. I'm glad you asked about this. So Spotify was an early adopter of Facebook Workplace, lovingly named Workface. I thought it was a very good tool for showing the true nature of a team because it was very easy for people to reflect reality in that tool. So I used it probably more than any other tool internally just to try to find out who is actually working with who. And then there's another problem that Stripe had to solve. It built its own tool for this. Stripe has a phenomenal, can we say intranet anymore? I don't know, an internal portal, something like that. Basically it was a tool specifically basically for this and I loved it. I haven't actually seen anything similar like that.Abi: And one question I have is how did Stripe keep this information up to date? I've talked to lots of organizations that have built internal tooling to capture this information, but the problem always seems to be, well, when do you actually update it? And typically they say, "Oh, we sometimes update it before an all hands or a company presentation." Did Stripe have any sort of mechanism for making sure this was kept up to date? Was someone in charge of it or was it a bit ad hoc?Jeremiah: Yeah, so there's basically two separate systems. There's one, the official HR system, which only people in HR can make those changes. And some of those changes then do get automatically reflected in other internal tooling. But each manager was responsible for, you could edit your team and then if you needed to go a higher level up to do something, you could move people to different places. But it was a fairly distributed thing in that regard. At least for keeping track of who is on which particular team. And then we used a completely separate process for team access controls and things like that. That actually for compliance reasons needed to be much more documented. So there was a process for doing that and then tracking those changes.Why the squad model became so popularAbi: That makes sense. Really interesting. Want to take a step back now and talk about the phenomenon, which was the Squad model, in terms of how it just really caught fire and spread across the industry. Seems like cargo culting is common pattern in our industry in general. We can look at things like microservices or the numerous new JavaScript frameworks that are constantly coming out and being adopted. So Squad model is similar to these in that Spotify published this article and you fast forward several years and so many organizations are jumping on the bandwagon and adopting it. So I'd love to just have a conversation about this. Why do you think the Squad model became so popular after it was announced?Jeremiah: I have a lot of theories on this. I think for starters, I think Spotify is one of the coolest companies in human history. So I think people are just naturally curious and drawn to anything it talks about. It's also a widely successful tech company and that success brings people who are interested in wanting to know more about what might possibly help make that company successful.But I also think that there's a lot of companies, especially in more established industries, that are undergoing digital transformation and they feel threatened and they're looking for quick fixes. The Spotify model is almost unintentionally a great misinformation campaign. It's deceivingly simple and it says things that people want to believe are true, like autonomy is important. So when it starts to then prescribe the other things, they feel like they must also be good ideas. Everything you read about it is super compelling and you really want to believe there is a simple, correct solution to very complex nuanced problems. So I think that's part of the allure.Abi: Yeah, I would largely agree with that. It does seem like every time a Google or a Spotify or Netflix comes out with some new way they're approaching technology or management... For example, Netflix Culture Doc, that was something that really caught fire as well. I think engineering is still such a new discipline in the context of human history. And so when something new comes out from someone who's been very successful, I think naturally a lot of people pay attention to it. And to your point, those tools that are published by the big companies aren't always things that should be adopted wholly by every organization out there, but maybe just more studied to see what bits and pieces of those frameworks or approaches can be integrated into what they're already doing.Jeremiah: Absolutely. I think there are multiple biases at play because we assume that what those companies do must be the reason that they are successful businesses, but unfortunately, most businesses are not differentiated in the marketplace because of their tech stack or their team structure or their culture doc. Those are just rarely competitive advantages for a business. And I think that can be hard for engineers who get excited about technology or Agile coaches who get excited about org structures and practices to understand. So it's fine to be excited about those things, but you all just also have to balance that by understanding that those are rarely areas of innovation that will differentiate a business.What the original authors may have felt about the popularity of the modelAbi: It's really interesting you included some quotes in your article from the co-authors of the Spotify model. I'll just read little bits of them out loud. One of the authors wrote, "Even at the time we wrote it, we weren't doing it. It was part ambition, part approximation." And another author wrote, "It worries me when people look at what we do and think it's a framework they can just copy and implement." I'd love to know, do you know more about the original authors feelings and thoughts on what has transpired with the Squad model?Jeremiah: Yeah. I do quote multiple other people who worked at Spotify and worked at Spotify at the time when these docs came out, Andesh Sheveson, he's actually somebody that I worked very closely with at Spotify. So I was actually really excited to get to meet him after I joined and to talk about some of the thinking behind the original white paper and how it evolved and played out. And then the other Agile coaches, I think both of them came after that, but have both worked at the company for quite a while. All three of them have given multiple conference talks and podcast interviews and other things, trying to explain reality to people. Like, "It's great that you want to be Agile, it's great that you recognize these particular problems, but this was really something about documenting Spotify's thinking at a specific point in time in its history." It was working a bit more in the open when it did that, and then it got really busy changing the music streaming business and it didn't really tell people about how it was continuing to evolve and what things actually worked.It actually came as a great surprise to Andesh when he was going to a conference and people were like, "We're using the Spotify model." And he was like, "Oh, you are?" And they're like, "Yeah, we just rolled it out to thousands of people who were even bigger than Spotify at the time." I think it just had much greater reach than, at least, Andesh ever imagined. And even today, I think the only people who are still really promoting the idea are people who are now running consultancies that are tied to the positive sentiment to it. But I think a lot of the other people who continue to work at Spotify have tried to be more truthful about how things evolved and what things continue to not work. And that self-reflection is really great. Unfortunately, the truth is harder to spread than an idea people really want to believe in, and that's part of the challenge.‍Abi: Well, I think this is a pattern that's repeated itself many times in the history of software development industry and will continue to repeat itself. So at the very least, it's really great to be able to get the ground truth, perspective, and history in this conversation from you today. Jeremiah, thanks so much for coming on the show. I think this is going to be so valuable to leaders and listeners out there who are wrestling with the problem of org design. Really appreciated this conversation. Thanks again.Jeremiah: Thank you so much for inviting me to be able to set the record straight and hopefully encourage people to keep studying organization design. There are no simple answers. These are questions that are worthy of looking at the past, seeing what works, and then hopefully finding something that works better for you.Abi: Thanks so much, Jeremiah.

This week on Transmissions: James McNew of Yo La Tengo and Dump. For decades now, he’s been a prolific source of engaged independent rock music—the kind we like here at Aquarium Drunkard. As past work like I Am Not Afraid of You and I Will Beat Your Ass proves, YLT are masters of a great sardonic album title, and on February 10th, the band continues that tradition with its 16th album, This Stupid World.   Transmissions :: James McNew (Yo La Tengo, Dump) Despite its title, This Stupid World is no doomer epic—grappling with the sorry state of affairs is only half of it. Digging deep, YLT turns up a mediation on time—and what we make with it.  “I see clearly how it ends / I see the moon rise as the sun descends,” Ira Kaplan sings in the opener, indicating the natural cycles that turn in the record’s fuzzed out epics. When McNew and host Jason P. Woodbury connected, Yo La Tengo had recently finished its annual Hanukkah celebration, which is where we pick up our talk. But from there, the conversation roves into interesting places: McNew’s dalliances with hip-hop, important Dump anniversaries—including the 25th anniversary of his Prince covers album That Skinny Motherfucker With The High Voice? From Yoko Ono to Sun Ra to the Dave Matthews Band, plenty of surprises pop up in this conversation—just like the YLT discography.  Thanks for checking out Transmissions. We’d love it if you left a review and five star rating on Apple Podcasts. Click the subscribe button so you never miss an episode.  If you dig the show, please consider leaving a five star rating or a review—or just forwarding your favorite episodes to a friend. We’re a part of the Talkhouse Podcast Network. Aquarium Drunkard is powered by its Patrons—if you’d like to become one, visit us on Patreon. Transmissions is written, produced, and hosted by Jason P. Woodbury. Our audio is edited by Andrew Horton. Our show is executive produced by Justin Gage. We’re part of the Talkhouse Podcast Network. We’ll be back next week with James Yorkston and Nina Persson of The Cardigans. For heads, by heads. Aquarium Drunkard is powered by our patrons. Keep the servers humming and help us continue doing it by pledging your support via our Patreon page. .

As research changes; becoming more data and computationally intensive, it is necessary for researchers to possess strong data management, data information literacy and data science skills. Research Data Service units in University Libraries themselves and in collaboration with disciplinary experts are well suited to offer training that bridges the gap between theoretical knowledge and application-specific or programming skills offering researchers the knowledge and how to to engage in this data and computationally intensive research environment. In this episode, Karl Benedict, a Professor, the Director of the Research Data Services and Information Technology Services programs, and a subject area expert for geospatial data and technologies with the College of University Libraries & Learning Sciences at the University of New Mexico, and Jon Wheeler, the Data Curation Librarian with the University of New Mexico Libraries, will discuss how the Research Data Services team at the University of New Mexico Libraries built a training portfolio to meet the data information literacy and data science needs of their research community. They will highlight how their training portfolio grew to include a variety of technical workshop topics, the diversity of methods they use to deliver these trainings, and how they assess success of these engagement opportunities.

Dr. Karl Benedict (@kbene) has worked since 1986 in parallel tracks of information management, geospatial information technology and archaeology. Within the College of University Libraries & Learning Sciences at the University of New Mexico (UNM) he serves as a Professor, as the Director of the Research Data Services (RDS) and Information Technology Services (ITS) programs, and as subject area expert for geospatial data and technologies. His previous experience includes fifteen years at UNM's Earth Data Analysis Center  (including five years as the EDAC Director), and work for the US Forest Service, National Park Service, and in the private sector conducting archaeological research, developing geospatial databases, performing geospatial and statistical analyses, and developing web-based information delivery applications. In these positions he has developed and managed the development of information technology and data management capacity in support of multiple research and application domains including public health, resource management, hydro-climate research, atmospheric modeling, disaster planning and mitigation, and renewable energy research. Dr. Benedict has translated this experience into both credit courses and workshops that are focused on skill building in data management and analysis and online application development, and his current work as a Carpentries instructor and instructor trainer.

Jon Wheeler is the Data Curation Librarian at the University of New Mexico, where he assists researchers with data management planning, curation, and archiving. As senior personnel on the NSF funded NM EPSCoR SMART Grid Center, Jon coordinates a statewide workforce development initiative focused on building capacity for data science education.

Resources Mentioned

The link to the Github repository where Karl and Jon develop and share their workshop materials is available here: https://github.com/unmrds. 

Information about the Carpentries can be found here: https://carpentries.org/. 

Information about the University of Arizona Research Data Services program can be found here: https://libguides.unm.edu/data

This presentation was recorded at GOTO Amsterdam 2022. #GOTOcon #gotoams

http://gotoams.nl

Adam Tornhill - Founder & CTO at CodeScene & Author of Several Books Including "Your Code as a Crime Scene"

ABSTRACT

Many codebases contain code that is overly complicated, hard to understand and hence expensive to change and evolve. Prioritizing technical debt is a hard problem as modern systems might have millions of lines of code and multiple development teams — no one has a holistic overview. In addition, there's always a trade-off between improving existing code versus adding new features so we need to use our time wisely.

What if we could mine the collective intelligence of all contributing programmers and start making decisions based on information from how the organization actually works with the code?

In this presentation, you'll see how easily obtained version-control data lets you uncover the behavior and patterns of the development organization. This language-neutral approach lets you prioritize the parts of your system that benefit the most from improvements so that you can balance short- and long-term goals guided by data. The specific examples are from real-world codebases like Android, the Linux Kernel, .Net Core Runtime and more.

This new perspective on software development will change how you view code.

In this talk, you'll learn:

• To prioritize technical debt in large-scale systems

• To balance the trade-off between improving existing code versus adding new features

• To measure long time trends in technical debt [...]

TIMECODES

00:00 Intro

00:33 Lehman's "laws" of software evolution

03:17 Technical debt

07:39 Case study: Prioritizing technical debt

10:25 Can we measure "code complexity"?

19:27 Technical debt & people

22:33 Case study: How quickly can you turn your current codebase into legacy code?

27:48 Resources

Download slides and read the full abstract here:

https://gotoams.nl/2022/sessions/1609/prioritizing-technical-debt-as-if-time-and-money-matters

RECOMMENDED BOOKS

Adam Tornhill • Software Design X-Rays • https://amzn.to/3DEeEnI

Adam Tornhill • Your Code as a Crime Scene • https://amzn.to/3FI5E2V

Adam Tornhill • Lisp for the Web • https://leanpub.com/lispweb

Adam Tornhill • Patterns in C • https://leanpub.com/patternsinc

https://twitter.com/GOTOcon

https://www.linkedin.com/company/goto-

https://www.facebook.com/GOTOConferences

#Legacy #LegacyCode #DeveloperProductivity #CodeComplexity #Complexity #Teams #TechnicalDebt #AdamTornhill

Looking for a unique learning experience?

Attend the next GOTO conference near you! Get your ticket at https://gotopia.tech

Sign up for updates and specials at https://gotopia.tech/newsletter

SUBSCRIBE TO OUR CHANNEL - new videos posted almost daily.

https://www.youtube.com/user/GotoConferences/?sub_confirmation=1

Luke Gearing interviews Marc Miller of Traveller fame!

Erlang & Elixir Factory San Francisco is now Code BEAM SF!

15-16 March 2018. Get tickets now →→ http://bit.ly/2nlioFp

---

Erlang & Elixir Factory SF 2017

http://www.erlang-factory.com/sfbay2017/larry-wall.html