#3 | On Openness | Interview with Kevin Marks

Ciao a tutti! Welcome to Engines of Creation podcast. I'm your host, Cristian Mastrodonato. In this podcast, I bring together my knowledge in complex systems with my experience in managing technological innovation and new product development to explore how successful products, organizations, and ideas emerge.

On today's episodes, we're going to meet a real web pioneer, Kevin Marks. If you don't know him, I suggest you look at the Wikipedia page dedicated to him, because, yes, the guests of this podcast do have their own Wikipedia page. Kevin is a software engineering expert who worked in companies like Apple, Google and Salesforce and have been at the forefront of digital innovation from the 90s. He brings a look that spans the whole horizon of how the digital industry has evolved.

and can evolve in the future. During our conversation we'll discuss how abstraction layers can commodify industry sectors and how this evolution towards commodities can be described using tools like worldly maps, but also our looking into what didn't work can help making sense of the complex fractals of these nested abstraction layers.

In the second part of our conversation, we'll turn our attention towards open models. Looking at examples in the web, like microformats, as well as in the larger digital industry, how open source set the standards for distributed governance. And we will close talking about what the approaches from companies like Apple, Google and Microsoft can tell us about how to find the right balance. This interview has been recorded before Elon Musk's Twitter takeover and Maurice's Generative AI explosion.

But I believe the discussion is relevant and timeless, spanning a large time horizon, and easy connections can be made to these more recent evolutions. This is the philosophy of Engines of Creation podcast. We won't rush to chase the latest trend, but we will spend... time reflecting on the patterns that create those strands giving you the tools to understand them.

Anyway, if you have any question about how the top is covered by the episode related to recent development or any other questions and feedback, of course, feel free to reach out at enginesofcreation at mastodonato.co or direct message me on my LinkedIn or Twitter profiles. both Sima Strodonato. I'm sure we'll cover more about these recent topics in future episodes, so hit the follow button. Now it's time to dive into the fascinating conversation with Kevin Marks, so let's go!

Welcome, Kevin. Thanks a lot. Thanks for joining the Engines of Creation podcast. It's a great pleasure to have you here today. It's good to be here. You are, in a way, definitely famous, but I think it would be worth if you can start introducing yourself, at least for the listener who don't know you already. Well, I'm a software engineer primarily. I've worked at a lot of different companies over the years. I was originally very much involved with media, so I worked for the BBC first.

did a startup called Multimedia Corporation in the UK in 1990, which built digital media, CD-ROMs, interactive video discs, things like that. I then went to California in 98, worked for Apple Computer on QuickTime on digital video there, left that for a startup called Technorati, which was indexing weblogs and making sense of weblogs. After that, I went to Google and worked there on social networking standards, OpenSocial. Then after that, I've worked for BT and Salesforce as...

vice president in charge of webstand type stuff which was less producing stuff and more encouraging the company to do things about it. And since then I've been doing freelance engineering for lots of different places including Digital Catapult and some for the BBC again. And also been very active in what's called the IndieWeb movement, encouraging people to have their own websites again and not spend all their time inside Facebook.

I think it's fair to share here a small anecdote. When we met the first time, I had the former CTO of Digital Caterpillar, Marco Balabanovich, telling me, oh, there is this Kevin guy. He's a great guy. You should speak to him. And I didn't know you. So I said, OK, yeah, that's fine. I need to have the time to look you up on the web and stuff. We sat down and you started telling your story. Wow, that's a colleague of mine. That's really cool. And actually...

To be fair, you also have a Wikipedia page. I cannot say I know many people who have their own Wikipedia page. And that I found out after we met the first time. So in that sense, it would be probably also worth for the listeners to have a look to your Wikipedia page. sense, you had an impact on the industry. And I think that the reason that I wanted to have this chat with you today. So the question where I would like to start.

is really about your extensive experience about innovation in the sense that you, through your career, you've been in a lot of places where new things were created and built and delivered and tested into the real world. So you definitely know how innovation works from...

from the inside. You told me a story about QuickTime was built, for example, which probably some younger listeners might not even know what it is. At least from my side, it was a big deal. I used it a lot and other things like software like that. So the question is really, what's your experience about innovation trends? What are the general patterns and trends that you've seen?

across the years that usually tend to come over and over again and you're like, okay, this is something that I know you should look for. Well, I think a lot of it is it moves in cycles. Each time there's a motion from new stuff that we don't understand yet through to products, through to commodities, through to standards.

So if we take digital audio and video as an example of that, in 1990, when we were doing the startup, doing that at all, we were still learning how to do it. We were literally... trying to come up with how you would get a computer to present video. It wasn't able to do it at the time. And even audio was extra work. Not every computer had a sound card. Not every computer had a CD audio drive.

At that point, it was very much, we're doing this from scratch. We can't assume there's anything there. And then gradually that became, the operating systems got more uniform on that. QuickTime was Apple software for doing audio and video. And they did a good job. I can say this because he did this before I joined, but he did a very good job of abstracting that in such a way that you could say there's different kinds of audio and video, this is how you play it back, this is how you capture it.

We can understand different file formats. We can make sense of the complexity of it, which means that as a user, as a programmer, you could just say, I want to open a movie and play it, and that would work. And that was...

two lines of code, and we'll give you a little box on the screen with a slider that moves and things like that. And they would deal with the abstractions of that. So then they sort of encapsulated the abstraction, which meant you could think about how does this fit into the rest of what I'm doing.

So one of the things we did, we used QuickTime movies as a place to store tiles for a globe so you could turn a globe around and look at it, which wasn't what it was designed to do, but because they had a system that let us go to any frame. in a movie quickly and draw it, we could just use that as an indexable image store. So you can take an existing thing and say, what can we build on top of that? What's the next piece we use that for?

Gradually, that set of abstractions become standardised. Now, audio and video are standard parts of the web. You just do an audio tag, put a file in, it plays. The browser deals with enough to think about it much. Or if you want to do something more subtle... There's WebRTC, so there are a dozen have-a-conversation startups at the moment, which are all basically built on that same stack. And they're able to say they're focusing on the...

How do we get people into a room and how does that work? Because the having a conversation over the web part is solved. We're having a conversation on the web now. We've got six apps to choose from to do it. And they all basically do the same thing.

because that's all been abstracted and shrunk down. So we get to the, each of these innovation things goes through that kind of path. So with any of these, you have to say, which point on the path is the thing I'm trying to do? Am I making something nobody's done before?

And within that, which pieces can I break into components so I can reuse existing things to build something new on top of, with the assumption that gradually that will go along the same bar, become a product, become a standard over time. And if you get it right, that's what happens. There's a sort of the weird fear of being commoditized.

But actually, it's a different kind of business, but a commodity business is a great business because it means people understand it and you can get a price for it. You're not having to explain what you're doing every time you go into the room. Literally, in 1990, we'd be going to companies and saying, we can do...

audio and video on computers and they would say what would I want to do that for and why and and how and you know we would have an hour or so of having the explanation and trying to come up with you're a wine merchant so we could do a thing that explains about different kinds of wines or whatever. Whereas now that's an assumed thing. People understand that audio and video is available to everyone. And we have multiple layers on top of that.

people are less concerned about how that works and much more about the actual stories they're telling and the different kinds of tricks and things they can do and some of the kinds of editing and composting that people do with it now. small children do kinds of digital composting stuff that was, you know, with the time I started out 30 years ago, was we'd be hiring people for two days to do that kind of stuff.

And so there is this sort of subsumption of stuff into technology that then gets fed into the next layer up. So with innovation, it's always you always want to think about what is the thing I'm trying to do? Something that is on the. the brand new side am i trying to compete with a set of existing products or am i too far over to the other side and it's a standard and i should spend time having talking to people about standards

and adoption rather than trying to build it. So there's always a spectrum of these. And what you find is that each of these produces a layer of stuff that you can build on for the next piece of it. So you don't have to write your own networking software anymore. You don't have to write your audio software anymore. So doing audio networking is easy. Whereas when I started out we had to write both.

It's a very interesting insight. What I'm thinking in this moment, it draws a lot into the concept of worldly maps. Yes. I think it's relevant maybe to point some listeners if they're interested into the worldly maps, because it's a nice way to map and understand what you're working on.

the product you're working on and in which phase are you really at the beginning of the process? So you're building everything from scratch. We are more towards the end that you are trying to sell a commodity. As you mentioned, there is nothing.

bad about that while sometimes you feel people shy away from commodities, but the innovation is somewhere else there. What you're trying to do is automate as much as possible. That's the thing. It's a different kind of innovation. And I think that's what we want to build there. Yeah. And that can be a kind of... If they're already potentially producers of this, but they're not commodified yet, you can work to commodify them. That's the Amazon approach, if you like, or some of the...

delivery service type approaches. You can say, okay, there are lots of places creating goods, but delivering them is not straightforward. We can take that piece and commodify that. And then you can say, It's a question of, is there a monopoly or are you trying to create fungibility in effect? Are you trying to create it such that you can substitute lots of different pieces in? In general, having a commodity, having substitutability is a good thing.

both for you as a consumer who's trying to decide what to buy, because you have more than one choice, but also for a business. it's often good to have an existing market that you can drop into so you don't have to build everything. So if you want to become...

supplier of a certain kind of thing, you want to be a bicycle manufacturer, there are a lot of component pieces you can bring into that to manufacture your own bicycles. You don't have to make wheels from scratch or anything like that. And it's a sense of what are the...

And the Wardley-Map point of view, as Wardley says, is everyone always draws diagrams of how the different pieces connect together. But the directions don't mean anything on their diagrams normally. They're just space. They're just trying to join things up. Whereas what he says, if you make the axes mean something, if you make one of the axes go from brand newness to commodity-ness along that way, and then the other axes he makes, how visible is it to the customer?

then if you start putting the pieces of your solution on that map, that's a useful way of thinking about it because you want to make sure that you're not building from scratch things you can get off the shelf because that's a waste of time.

And you want to be sure that the pieces the customer can see don't change too much, even if you can change the pieces underneath them as well. Or alternatively, if you're trying to surprise and delight the customer, you want to say, this will be a visible change.

If you're arguing about the details of some technical implementation that they don't care about, that's not something that you wouldn't be able to explain to them. So you've got to show that it makes sense at their level as well. So it's a very good framework for understanding the...

understanding what you're making, but also communicating with other people you're working with which bits are which. And then Wardley goes into much more detail of this, but one of the key things he says is different kinds of work practices are appropriate to different parts of the map. And if you're in the...

Commodity side of the map, there's very much a different work process than if you're in the experimental side of the map. Because in the commodity side of the map, you really care about uptime, consistency, service levels, availability, and that kind of stuff.

you're probably better off buying that kind of stuff from a large provider because they can focus on that rather than trying to build that from scratch yourself. Or if you are, you need to understand that's what you're doing and you have to have more of a Six Sigma approach or that kind of...

high attention detail approach. Whereas if you're experimenting, you don't want to spend six months on the planning document, you want to build some things quickly and see what works, and be able to iterate fast on that until you're comfortable with something you can then make the product story, and build that with a more assembly point of view. So basically, he's got...

he's been thinking about this a while and he's written this stuff up quite well. So I recommend reading really broadly on this. And that was very illuminating for me when I did that because I've obviously worked different parts of that cycle as well. You know, when you're working on operating systems at Apple, you're not...

often in the experimental phase, because anything you ship, you've then got to support for the next 10 years because people have linked their binaries against it. There's a certain hesitancy before you ship stuff at that point. We can experiment and try things out.

But at the point you say this in the operating system, you're making a large commitment for the company. And that can be the same for anyone who's providing APIs to other people is effectively taking that on. And if they're not, then they lose trust. And we've seen that recently.

with some companies suddenly say, oh, we're not doing that anymore. And you're like, I was relying on that and now I can't use it anymore. Makes sense. A very good point. And I think it touches probably with other topics. around the influence of time into processes that often is neglected. And we had some discussion about ergodicity, maybe we can get back to that. But in this moment, more than focusing on time, one thing I'd like to ask you is,

What do you see the trends now? Okay, so there is a lot of talking of the moon's low and as a major trend. It's influencing all our world for the last... 30, 40 years in a way, we can see that in operation. Where do you see there are now other trends that are coming up on top of the more low that are influencing more and more innovation and how the market are evolving?

Well, so, I mean, Moore's Law is sort of reaching the point of diminishing returns, let's put it that way. In order to go down the next level of density in fabrication, the amount of money you have to spend on it is... um nation state level which is why there are only two or three fabs that work at that scale now but what has changed is the the original formulation of moore's law was based on number of transistors on a chip and it's since been basically modeled on how much faster

single processors are getting. But the trend has been instead to put multiple processors on chip and to work in that direction. So if we look at CPUs, it's gone from how fast is a single CPU to effectively GPU-like cards, where we have large arrays of processors that are doing things like that. And then the other trend has been to reduce the amount of power, so you can do the same amount of computation with less power, which is why we have mobile phones.

more than we have laptops now and you can see that trend extending where suddenly the next generation of Apple computers are being sold on the battery last two days rather than it's faster than it was before. Somewhat faster than it was before, but the primary thing is they've picked up the mobile phone worldview into the chipsets there. Then the other half of that is, rather than trying to put everything in one computer, you say, how can we use lots of computers at once to do something?

which has been the distributed systems thing, the sort of Google data center approach to computation and big data and those kinds of things. That has been a different kind of software architecture. So part of the trend there has been from, if you like, the classic sort of von Neumann architecture where you have a CPU and memory and you build things that way to something that is...

closer to a functional architecture model where you're calling things, you're not quite sure where the things are coming from, and when they complete, you wire them out together again. And we have had some success with building new models that work in that way.

But I think we're not all the way through that transition yet. And I think one of the challenges is expressing that in a way that makes sense to people. It is still quite hard. And if we look at the way... being programmed models, things are a lot more asynchronous and there's a lot more of that sort of worldview in there but we're still often trying to write it in ways that look synchronous.

to the weird point where the new JavaScript model is, you put async in front of something and then call it as if it was synchronous and things are supposed to work for you, which it does until you get an error and then you're not quite sure what's happened because the error cases are all hidden away from you. There's a sort of...

assemblage and a path dependence thing going on there that has sort of crept up on us such that everything you touch is connected to everything else. And part of that is the kind of useful... encapsulation idea i was talking about before we say i don't want to think about how to play video i just want to play a video here what i care about the comments on the video or wherever it is and part of that is there's a bunch of

sort of abstractions that anything you touch is relying on and they've all got to be working for it to work. And that is the ongoing tension of digital life at the moment. So what I'm seeing is there's lots of... conceptual churn going on where people are talking about one of this first microservices big data deploying functions run deploying applications serverlessness and all these kinds of things and these are all sort of

different attempts to try and say how do we abstract this much more complex thing how do we slice it into smaller pieces that make sense to people and we're still you know we're in the middle of that we can see patterns in that but it's It's not as straightforward as it used to be. And then the other half of that is we have the, because we've got, you know, Turing machines and emulation, you can actually nest these things inside each other.

So one of the ways you deploy software now is you take some existing software, then you compile it into WebAssembler and use it in the browser. So I can boot my CD-ROMs from 30 years ago in the browser. in an emulator and they'll behave just as they did 30 years ago because it thinks it's running on them back into us from um 1994 and there's this there's sort of layers and layers of that as well so we started end up having sort of nesting dolls

of abstractions going on it can take a long time to unpick what's actually happening so there's there's the idea that things are simpler now because we know how to do certain kinds of things but in order to do that there there is a sort of fractal nest of complexity inside it Thanks a lot for this explanation, first of all, because it was really illuminating. And I was really thinking about complexities all about that. And it's really about having different actors and agents that sort of...

trying to play together in some way. And it can be random, apparently, or can look like random. But as you can see, in the end... Patterns emerge naturally, which is the new level of abstraction, but it's difficult to predict them. You need to play along and see what happens. And you can guess, but you know only when.

when it happens, when you know that? Is it fair? Or do you think is there a way to at least have reasonable guesses of what's going to happen next? I think it was easier for a while because we had the morsel thing going on, so we could extrapolate that.

There was a sort of feed-forward effect where also the chip people were looking at more and deciding that's what their schedule for making things smaller was. So you could extrapolate what they were going to be charging for things next year as well. In a sense, we've run out of that space.

You can keep making computers smaller, but there's an irreducible utility to how you interact with them. So if you make them too small, then you've still got to work out how you talk to the room full of little computers that you've got lying around doing things. And we're hitting that.

that phase at the moment we um the number of network devices we have in the house is growing at a rate where it's like i have actually no idea what these what these things are and what's going on and that's fine until one of them starts continuously broadcasting data and knocks everything else out. And you're like, what's going on? Where's everything? There was a point where suddenly everything was getting internet connected because it could be.

Because you can buy a chip that can do wireless networking and power it and attach it to whatever, then suddenly everything decided it had to have a chip in it and would do that. And we end up with lots and lots of flaky gadgets that... have chips in for no apparent reason. There's no real reason for my thermometer to be network connected, but it tends to be by default. And then suddenly we've got a large...

amount of sort of wasted computation doing weird stuff and chewing up network bandwidth so there's there's like there is that sort of overshoot going on and then the other part of that is that you know

People are looking for what is the next thing that's going to make sense? And part of the challenge there is that that's been very financialized because investing in technology has been a... a useful way to make money for 20 years or so that's sort of become its own the thing in itself so a lot of the rush we're seeing into cryptocurrency type stuff and things like that is a sort of explosion of financialization beyond the point where it actually makes sense because because people understand

Well, because people in technology understand what stock is because they've been paid in stock for long enough, they're excited by Bitcoin because it feels like a stock. Even though it's pure speculation, there's actually very little utility there. And there is this...

There's that sort of sense that the other half of the technology excitement is there's always a lot of excitement about what's the next big thing going to be, but we forget what the supposed next big things that didn't happen were because they didn't happen.

And there's lots of those that burn out. And so in the moment, if I look back 30 years, I can say, well, obviously I was on the right track with audio and video stuff, but what were the things we were trying to do in 1990 that didn't work? And there was... a lot of effort there to try and do rules-based expert systems, for example, that people put a lot of effort into and not much of that came out. I mean, effectively, they were...

they were trying to build an abstraction that was more useful than a relational database, and generally it wasn't. And we've seen a lot of that with a lot of similar things. Well, I mean, there's been a great excitement around data science, but in practice... a lot of that is, well, it's just like trying to find where the databases are in the company and wiring them up. That's a useful thing to do, but the amount of actual science and stats we're doing there is often fairly small.

So sometimes the hype runs a little bit ahead of the systems. And the machine learning is like the step beyond that. We have had some great breakthroughs in machine learning, especially on the image recognition side of things. And that has been exciting. But that depends on that layer before of actually...

having useful data for the machine to spend its time on. The fact that we have millions of photographs being taken every second means there is raw material for stuff to happen there. So we can have a...

a machine learning model for doing things there that were hard to do before. But it doesn't necessarily mean that it's useful everywhere else as well. Also, the other big thing is, what is the model missing? Whatever you're modeling... is always an imperfect model and you're usually blind to what the model is missing because you can't see it in the model and that is that you know if there's one there's one thing i'd say about technology is

Learn to look for the thing your model isn't covering because it's not going to tell you about it. You're not going to see it. Most of the time, your code isn't going to be there. And that is often the place where either... you know, the whole thing will come tumbling down because of a fault or you will actually not be solving the problem you thought you were solving because you can't see the holes in the gaps. Okay.

Let me touch the point from AI just because I wanted to close your thought, but then I would like to move on on another piece that you mentioned that I think is really relevant. So regarding AI, I agree with you. It's really interesting. And I think there is an element where... Sometimes, and it also, technologies are tried too early, and we are talking about expert systems. Now, what you could see is that while there is a big deal about statistical models in today's AI world, you can also...

There is actually enough computational power to start talking about extra systems again. And you can hear and hear more talking about hybrid models. So there is always, again, time is an important variable. It might be, is the right technology at the wrong time? And there are a lot of situations where it's kind of...

come back again. And I'm using this specific example because I would like to talk with you about something that I think is particularly relevant nowadays, which is... your experience actually around the first attempt so on the social media space and the work you lead around open social and micro formats okay which i think is really valuable i think at the time and uh you're probably talking

Early 2010, something like that. A bit before, yeah, yeah, 2000. It was at the beginning of the social media. And then probably the dream didn't really come true at the time. But... And that's what I would like to understand from you is that I feel it's coming back because it kind of became a monopoly, but we also can see now what are the consequences. I started talking about, again, having open formats around social media.

this sort of thing. So what do you think about that? I mean, it might be the right time? Well, I hope so. I mean, I'm interested in it. So the challenge with this is... It is the gap between monopolization and commoditization. And one of the things that you try and do if you're not the one with the monopoly is try and encourage the commodification.

So that's part of, you know, that ends up being one of the ways battles of large company type thing. So I was involved in Google's attempt to commoditize social networking, but I believed in it because I'd seen what we'd been able to do with. the web and with blogging and with microformers and so on. And so we had a lot of models there for how you could get multiple different sites to communicate through standards. The basic one obviously was HTML.

in url so people go to url and read something but then the microformat's point of view was to say well okay within that html we can add a bit more metadata a bit more information about what's on the page so that we can crawl it make more sense of it so the technology companies i worked

with the technology was we were trying to keep track of all the blog posts as they were made and see who was linking to whom so that we could tell, we could spot the conversations going on effectively. And that meant having to extract structure from the blogs, but also encouraging them to publish more structure and create a feedback loop to do that. And so we were able to do that with a combination of existing feed specifications.

and microformats embedded in that to add a bit more information. So we could say, this is a tag, this is a like, that kind of thing in there and extract that out. And that still works. But the challenge with that is that... Doing that across multiple sites is inherently more of a coordination problem than doing it within a single site. So what we saw was that the larger sites...

Well, what became the larger sites like Twitter and Facebook and MySpace originally would adopt these things early on and be part of the ecosystem. And then at some point they decide, actually, we just want to pull that inside our own database and do it our way instead. And we've seen that happen.

over the years with all of them, really. Jack at Twitter is saying, oh, we should do some more standard stuff like we used to do. It's like, well, why did you stop? Twitter's original model was very much they build the API first.

and release the API at the same point they're building the site so other people can build things on top of it too. And they broke with that after calling a large developer conference and realizing that other people were going to compete with them on different kinds of clients and things.

and they still hadn't worked out where the money was in Twitter, and they were very worried that someone else would find it and take it away from them. But what that meant was that their ability to be an ecosystem... shrank radically when they took that stuff away. You can still do quite a lot with Twitter as an ecosystem. There is some API there, but it's much less than it was 10 years ago.

And similarly, Facebook, again, they did a similar thing. Facebook, when we were doing the open social stuff, we were saying, okay, we've got 50 different social network sites. They've all got basically the same thing going on. You've got profile pages. You've got posts you make.

and you've got friend connections. So we abstracted that. We said, okay, this is what a profile looks like. We looked at 20 sites and looked at the fields in the profiles and collated them. This is what messages look like. This is what the feed looks like. Abstracted that. specified it and then built a wrapper around that so that you could basically write the same app against multiple social networks. And that's a promise and a threat. I mean, and the social networks were like...

Okay, sure, we'll work with you on that for a while. And then they're like, but ours is special because of X and ours is special because of Y. And it's like... No, yours is special because everyone on your network speaks German and yours is special because everyone on your network speaks Spanish. And that's why you exist separately because that's the community you've grown into. But in terms of what people are doing, they were mostly fairly similar.

And eventually Facebook was sort of able to take advantage of that by saying, well, we're the, you know, we're the shiny rich persons network. You should be on ours rather than on your weird parochial Spanish speaking network. Come and join ours instead. And they brought people in that way.

And also by being, this was their entire business, they were extremely focused on that. Whereas trying to do it by coordinating 20 different social networks in different companies didn't work as well. And so that was a... Definitely, there was a failure to standardize there, which meant that a monopoly took over. But Facebook was smart enough to provide the APIs that people wanted at the time so that people could build things on top of Facebook, and then they gradually withdrew those.

every time so at the point at the point in 2009 2010 where facebook were subsuming other people's code into their system they were very much Yes, build anything you like. You can connect to anything inside the system. That's fine. As long as you're growing Facebook, we're happy. And then they realized, oh, wait a minute. How about we own that instead?

And they pivoted to do that. And similarly, the point after I left Google, Google decided to say, oh, we're going to do that ourselves too. We're going to build our own silo that is all ours and subsumes everything. And that was what became Google+.

And then they gave up on that 10 years later, and they don't have that anymore. I mean, they've still got YouTube and Blogger and bits and pieces, but they don't have even internal cohesiveness that they had that we were trying to build inside Google of. a single profile account and connections between the different systems. So they managed to be less coherent than Facebook on that one. The IndieWeb part of this is, at the same time,

The web still exists. It never went away. We can still build our own websites. We can still write blogs. We can still tweet. We can make podcasts. We can all our stuff up that way. So the IndieWeb stuff is saying, well, let's keep doing what we were doing, but rather than try and do it by...

being Google and saying let's get everyone in a room and tell them how to do it. Let's build the smallest possible thing that can work to connect two sites together and work on that. Okay, once we've got that, let's build it as separate components and only build what's needed rather than trying to do this.

giant synthesis sort of top-down point of view so that that i think that is working on the indie web but it's it's working slowly we've been trying to build these pieces out and we've got we've got some stuff that works and people suddenly knows oh right yeah i want to comment on your site i can use a web mention

And that will go through. And someone's done the work to wild it up to Twitter. So if I comment on Twitter, that'll show up too. That's neat. And we have the infrastructure to do that across sites rather than within one site. But it's still, because it's components and pieces, it's more integration work.

And that's the trade-off. I mean, I suppose the counter example, if you like, is podcasting. Podcasting was very much an extension of blogging, and it took off. And it took off. Partly because of Apple, because the part in podcasting was because Apple said, well, what if we had things that played audio in our ears when we weren't at our computers? Spoken word onto that was a useful thing.

So we built that. But Apple was not concerned about that being their business. They were happy selling you iPods at the time and then iPhones later. And if... Part of the value of them is that they have access to audio. That's good. They built the music store, and that was initially... This is forgotten now, but when they first built the iTunes store, Apple was not making money on selling the songs because it was costing them as much to process the sales as it was to...

as they were charging the record companies for it. They were charging 30% then, but that was what it cost to do a credit card sale in 99-ish. No, yeah, 2001, I think it was. And so...

From their point of view, from the record label's point of view, that was fine. Apple was doing the messy work of selling the songs for them, and it was less than the record companies could do it for. And Apple didn't mind losing... losing roughly breaking even on that because they were selling their physical hardware which is where they made their money and that went through and the podcasting had they had the same attitude to that which meant that there wasn't a there wasn't a sort of

urge to push it all into one company in the same way and we've seen that that has that is you know there are several people have tried to do that since but podcast has sort of maintained this abstract, standardized thing. That's a good thing, but part of it is that it hasn't grown much because we're still effectively using the same file formats and models from 2004 or so when we've pushed all this together. And that's...

But it does mean that you don't have to pitch somebody to start a podcast. You just start a podcast, and then you end up pitching all your users to finish Apple for you. Yeah, that's a very good analogy. And I'm thinking of myself and know how to, let's say, the initial struggles on how to set up a podcast. But I see your point. But I think it's a nice way where...

Standard, in a way, worked, even if they are, let's say, as you mentioned, oldish. And I think, as you mentioned, it's been relevant in a way that probably nobody really wanted to have the control from day one. And that's the reason because maybe now after 15 years is having its success that now podcasting is...

It's exponentially growing over the last couple of years. And of course, definitely the COVID situation, of course, influenced that. But there was a platform there. There was a decentralized platform that could help anybody who wanted to, let's say, share.

in a more natural way. Probably more natural compared to YouTube, which is a different thing because you're a little bit more slave of the algorithm, okay? And there are other platforms. And I think the podcasting is a little bit more freer. That's what I think that...

And there are some probably similarities. I like the similarities between podcasting and, let's say, the indie web. I mean, the thing with podcasting is that it's… in some ways because there's a it's a bit more intimate because you're usually doing it listen to something you've got a voice in your ear and that that's even though you're you know you're doing that we're doing something else it's less sort of bitty than

interacting with Twitter or Facebook is because you're actually you're usually coming to a podcast for a bit longer period of time which means also you can't listen to that many of them so there is a sort of a lock-in effect and also a parasocial part to it. You subscribe to a podcast and then you get used to listening to those particular people and they take up one of your notional friend slots. In the same way that broadcast TV presenters and people do, it means that there is a...

It's harder to chop into small pieces and glue them back together again. We haven't seen much of that when we're podcasting. People have ended up with half-hour length things much more often than they have with five-minute length things because there is that sense of a comfort and a longer conversation.

and a certain amount of friction to what you play next and so on. So that, I mean, that may be a, you know, that could maybe change a bit. We could have shorter and tighter podcasts. Whereas, you know, video went the other way. Video was initially everything was short hits.

and you would just get little blips of things. YouTube used to have a five-minute limit, I think. Originally, maybe it was 10 minutes. And then the other platforms have gone with fairly short limits as well, if you look at TikTok and... Yeah, and Vine and the other ones that were successful, they went through very short and bitty and basically doing the tweet model of video. Whereas YouTube somehow went the other way. It was algorithmically driven, but they rewarded longer.

duration things and suddenly we have the sort of the half hour video essay is a is a sort of part of the sort of the youtube genre if you like It's being like quick hits and has become these sort of longer things that people actually share. As well as the, like, here's the funny thing that America's Home Video type people falling over and hurting themselves stuff, that's still there. But they...

Not sure if fully deliberately, but they did have an OKR to say we want more minutes listened to per person. And so we will work out how to do that. And part of that was amplifying longer things. Yeah. And it might be, of course. let me say it might be not of course it might be just my guess is that it might be related to

So the revenues, the more menus, the more commercials you can put in and the more revenues you can get that might be the relation there. But you can see with TikTok, you can build profitable models also going short. That OKR might have been based on that assumption that wasn't necessarily right. One thing I would like to pick up about one thing about your experience, because what I'm curious about, I'd like to have your point of view, which is related to the, let's say, distributed.

social, the open social experience. The distributed experience, which is sort of bottom up, and you try to play with complexity because you want to see new patterns emerge, but it takes time. It's not something else. It takes 10 years. It's going to take 15. We don't know. Versus, let's say, the top-down monopoly approach that Facebook had. And I think this is an interesting metaphor around many things that we are seeing around us in politics, in the way social, the overall lens.

is shaping around us. And you mentioned about, let's say, the worldly map. So, for example, it definitely is an element where something becomes a commodity, might need a little bit more control over it. Because otherwise, you have complete control on the market if you become a monopoly. And that could be the situation on the Facebook situation. Like, of course, a distributed governance model as a principle, it keeps the right level of freedom, probably, that we might need.

I'd like to hear your take on how distributed governance could play out. So, I mean, one of the big successes that we have had in a sense of distributed governance is open source software. So what happened... there is that we've gone from the default being software is proprietary and owned by companies and you never see it to most software that people are actually running is now open source and visible.

and forked and checked into visible source code systems. And that was a cultural shift. It was a cultural shift in engineering. I think it was somewhat engineering-driven, but it was a much larger cultural shift in management and company understanding. And part of that was that they realized they weren't saying software, they were saying services or they were saying physical objects. And so they were less worried about open source because they went.

charging for the software and that's why it took long for Microsoft to get it than others because they were charging for software. But that has been fairly successful at shifting governments from an adversarial patent and lawsuit driven worldview to a, we're all trying to use the same code and trying to change it. And there's sort of, if you like, there's sort of the biggest, the biggest cultural shift there, I think.

has been GitHub going from, you know, 20 years ago in open source. If you were going to fork software, that was a declaration of war on the maintainers. And this is even once open source had happened. the idea that you would try and make another version of it was you were obviously about to start a fight. Whereas the first thing you do whenever you touch a project on GitHub is you hit the fork button and change it and then have your own version.

publish your own version out there, and then you say, oh, if you want my version, here's a pull request. So they've taken, in effect, they've diffused the hostility of that into a different gesture. It was... intrinsically built you know the original underlying git stuff was built for linux for linus to manage all the people who were sending in patches and he is famously a grumpy person who will argue about you rudely about your patches but he built the

He built the ability to patch that. And then GitHub wrapped that in a different social convention of fork first, ask question afterwards. Because of that, that has changed the software ecosystem.

radically and the difference between linux and npm if you like is the difference between an open governance that still ended up with a single person in control and then actually distributed governance so that if you want a sort of cultural model for that that's that's one example another example to think about and this is one of the things i get frustrated with some of the people who talk about decentralization is they say we've got to get away from dns it's too centralized it's like well

resolution of DNS is centralized but actually the entire system is fungible and commoditized. I can buy a domain name from a thousand odd registrars and then the other rest of them will respect it. and the entire DNS system will respect that. And if I don't like that registrar, I can take my domain name to another registrar and another DNS provider and another web service provider and all the pieces up the stack. But by paying for that domain...

and part of an ecosystem that is fully commoditized in that respect. And it's a lucrative ecosystem. You know, there were Super Bowl adverts by people who say the main names. This isn't, you know, this isn't some small business. It's fairly large. That's the sort of alternative to Facebook controlling the namespace, which is what they do inside their system. And so they can throw Donald Trump off Facebook. They can't take away DonaldTrump.com from him easily.

Okay. They could take that on trump.com away from me because ICANN has a... dispute resolution mechanism for when you're taking other people's names and you could get to that point so that you know there is a sole problem for that but it is a distributed process it is a it is not a fiat process in the same respect I think the examples you gave are really interesting. And I think at the end it's about generosity. I like the point of GitHub and it's really about being open and say, yes.

you can take my work and it took time in a way it was a cultural shift I see that because even at the beginning the open source software was still about okay but that's my project that's yes that's our code it was a community code but still they were their code but now it's a different completely radical shift words. This code is really for everybody. So please make a copy of it and think about that. And then you can do whatever you want about that. Listen, I would have other...

Under the question. And to be fair, I had a lot of other questions I wanted to make you that probably it will be for another time. Let me close with a final question. What do you think a leader, and I'm using leader, of course, as a very wide term here, it's not a manager or a boss, but really somebody.

wants to make the difference. What do you think should be the most important characteristics nowadays for a leader? Well, yeah, there's different ways to lead is the other thing. I mean, the broader thing is... The idea of servant leadership is interesting. The idea to say the point of the leader is you're making things easier for the people you're trying to lead. And if you're not doing that, why would they follow you?

That worldview is sort of part and parcel of the same thing. There are alternatives to whatever you're trying to be able to do, so you need to provide them with a reason to do that rather than just gather them and try and point them in a particular direction. So that hasn't necessarily percolated through everywhere, but you do see signs of that in changes in political leadership and in...

And in sort of organizational leadership too, you see companies that are built less as a single center of truth. You know, Apple, when I worked at Apple, was very much that. Steve Jobs was the person that everyone had to pitch through to. And there were layers of pitches to get through the other people to get to pitch to Steve Jobs. And so there was this intense sense of...

You can try things, but it only actually happens if it gets through to Steven back. And there was part of a reason for that is that Apple was primarily making large physical objects that take several years to manufacture. And so they had to commit to things to do that.

But that sort of echoed into other practices as well. It's one of the reasons that Apple was, even though it adopted open source software, people don't talk about the Darwin kernel particularly. You know, it's a FreeBS default that Apple's stuff is built on, but it's not.

You know, it's there, it's subsumed, and Safari is open source, but it's WebKit and there's stuff down there. And it's there, they've sort of subsumed it's part of the mechanism, but culturally they weren't ready for that being part of the company. As opposed to Google's earlier point of view was very much, okay, we expect lots of stuff to be going on in parallel in this company, and that's a good thing, and we won't try and...

will encourage people to build things first and ask questions afterwards. Because they were primarily a software company, they built a bunch of infrastructure you could build software on quite easily. You could try things out and experiment and throw them away. And the drawback there has been...

there's now a sense that Google makes products and throws them away. So people will say, oh, Google's done a new thing. Okay, I don't care. They're going to cancel it next year because their experimentalism has been turned back on them from that point of view. With Microsoft, they've gone through that transition in an interesting way because they've gone from decentralized, shipping a single product, has to be supported for 20 years.

company to, oh, actually, we're trying to build lots of open stuff that runs in open source in the cloud and that kind of thing. And they're sort of part way through that process. But they're trying to thread the needle between the two models. And it's been interesting to watch them do that and to see that transformation. And they've still got a lot of that, you know, we are building the one true version inside, but they've...

But from the leadership point of view, they have said that we have to learn from everything else. They own GitHub now. They have subsumed the idea of development happening massively in parallel. They've distributed VS Code freely and made that open source. So having gone from developer tools being a profit center, that is now part of what they do as part of that ecosystem because they've realized they can get a lot from that ecosystem. So there's...

It's just to watch different companies adopt different parts of this and go in different directions.

Taking an example of Microsoft, it's worth noticing that I actually acquired GitHub some time ago, going in this direction. And GitHub is an example of a company that actually changed the way we look open. Yes, exactly. But I take it as a... at least personally as a lesson because it's also my personal lesson is that moving from close to open is definitely the trend and that's probably the we can see this trend and you can see the most successful company to be

more and more open. I think that's probably what modern leaders should look at. I really take that as a sort of a a core lesson through all this discussion we had today. Thanks a lot. It was a brilliant conversation. A lot of things to, I believe, for our listeners to think about and hopefully they will be able to. come back and make questions and maybe follow up afterwards. So thank you very much, Kevin. Thank you. Great to meet you.

Thanks for listening to this episode of Engines of Creation. I hope you enjoyed learning from Kevin on how innovation and complexity shape the digital industry. If you did and you want to learn more be sure to subscribe to Engines of Creation and leave us a rating and a review and if you have any questions comments or feedback feel free to reach out to enginesofcreation at mastronato.co or direct message me on my linkedin or twitter profile both si mastodonato arrivederci