Matt Mullenweg on Open Source, WordPress and Distributed Work

Matt Mullenweg is the co-founder of WordPress, the most popular publishing platform on the web, and the founder and CEO of Automattic, the company behind WordPress.com, WooCommerce and Jetpack. He is a longtime supporter of Creative Commons and the open movement, and CC Search, the search engine Creative Commons built for openly licensed content, found a new home and is now part of WordPress.org. The episode also covers the 2021 CC Global Summit, taking place virtually from September the 20th to the 24th.

He started with proprietary web scripts for forums, including vBulletin, and quickly migrated to the open source alternatives because he had no money. The alternatives he names are phpNuke, which he describes as a Slashdot clone in PHP, and phpBB, which he describes as a vBulletin clone or forums clone. As a kid he would pirate things or crack video games, but that never sat well with him, so he grew out of it early and began either paying for things or finding things that were meant to be free.

Creative Commons was incorporated in 2001, with the licenses first released at the end of 2002, and it was very prominent in the blogger community at the time. Matt followed people like Aaron Swartz, Lawrence Lessig and probably Joi Ito who were talking about Creative Commons, and he had a sticker on his laptop that said 'Free the mouse'. He says the Free the Mouse campaign, which Lawrence Lessig worked on and maybe Cory Doctorow as well, was maybe his first introduction to how copyright worked, along with extensions and the public domain. He always found it fascinating that for content, unlike software, there was not a great way to put things out there.

It is the world he wants to live in. He supports people's rights to build proprietary versions and sell them, but for what he personally works on and contributes to, he wants to increase the commons, the collective wealth of all humanity, and the ability and the access. He finds it exciting to see open source applied to areas other than software, whether that is human knowledge with Wikipedia or finance with Bitcoin and Ethereum, and he says this is what he plans to do with the rest of his life.

Creative Commons search, the open search, now lives on wordpress.org and they are hoping to build it out quite a bit, and he wants to make it easy for people within WordPress to license their images in a way that can be accessible to others. WordPress has always included design and themes, but it was almost impossible to find images they could say were GPL, since GPL is not really designed to apply to images. So he took the about 30,000 photographs on his site that he had taken himself and said to consider them all completely open, and the theme team can use any of them they want. That is why his photography appears in a lot of the early themes for WordPress.

Friends who work in film or music really appreciate the open licensed assets available there, and as a creator who does film and music and photography and code and writes a lot, he gets rewarded by seeing people experience that work and use it, and he loves when it gets remixed or modified. He notes that this is not right for everyone and he is not saying it is better than the alternatives, it is just what he wants there to be more of in the world.

It is still a challenge. On the engineering side there are decades of people contributing code to the commons, in nights and weekends or as their main job, but in design that is not as common. There are no issues with translation, which appears to be a very collaborative field, and no issues with volunteer support or documentation. But in areas like marketing or design, those communities may not have as much history of contributing to a community or commons project.

Matt thinks less about collaboration and more about community, and about blog comments in particular, and the idea that every blog post was actually the start of a conversation, not the end of one. The collective intelligence of the internet coming to your front door on whatever topic you wrote about was one of the most compelling things about blogging. Growing up in Houston, Texas, where the technology community passionate about open source and Linux and Wi-Fi was relatively small geographically, he could connect online with worldwide enthusiasts. He thinks a lot about how to reinvigorate that common space, because conversation has moved to social networks which dilute it, since their model is to get your attention and change your mind on something usually commercial. He has even noticed on Twitter recently that he is not seeing a lot of the replies to his tweets in the web interface, so he spun back up TweetBot, their desktop client, which appears to be a cleaner feed.

The concept is to combine a blog and a wiki so that websites powered by WordPress are not just one way outside of comments, for example if you saw a typo on a post you cannot just click an edit button and fix it. WordPress by default sends the first comment you make into a moderation queue, where it can be approved or denied. The idea is that a suggested edit to a page or post would go into a moderation queue, and you could approve it, deny it, or send an email, so it could be the start of a conversation not unlike submitting a pull request to code. Bringing an edit button to everything, kind of like everything on GitLab or GitHub has, could create more community-oriented knowledge sites that are specialized or in niches.

He says the experience of WordPress businesses is hopefully a counterexample to the perceived tension between open knowledge and sustainable businesses. All their code is open, and sometimes people take the code and resell it and compete with them, and that still happens to this day. But he has found that the more he has given away, the more he has gotten back. It forces you to be good, because you are not going to have a market just by being the only person with some feature or line of code, so you have to figure out what really distinguishes you and what value you are providing. The WordPress ecosystem has a strong community norm and license of being 100% GPL, which he describes as a very permissive license, and still drives, not counting e-commerce, $10, $12 billion of revenue per year across all the players there, which is his estimate.

He underappreciated the importance of marketing and sales early on. Because he had a background in web design and was very close to customers, they got some things right and lucky in terms of positioning and branding for WordPress, but he now really appreciates the wisdom and experience that people with a dedicated marketing or design background, or marketing and sales background, bring to the evolution of their products. For sales, it is like learning from your customers through a different lens. He resisted having any salespeople for the longest time at Automattic, and now it is a source of a ton of learnings about what real customers in the field are struggling with and what they want. He followed a misshapen idea that engineers can do everything and should do everything for many years, and later came to appreciate a specialized team with great experience coming together to work on a common goal.

They had already started work on an open directory, a place for people to upload things, so that was on their mind, and the opportunity was to not just have what people contribute directly on wordpress.org but actually index the World Wide Web of the great content people are putting out there. They can also tie it into WordPress itself. He loves systems being as distributed as possible, which is a big mantra for him: distributed companies, distributed work, distributed collaboration. The fact that people can distributively publish on their own websites content they want to be part of the commons is powerful, and a search engine, as Google has shown, is a really fantastic way to do that. A big challenge for the directory is how to surface what people are looking for when they search for images inside of WordPress, which is a little closer to stock photography.

Two open questions he names are how to navigate model releases in a Creative Commons way, and how to handle cases like the twinkles on the Eiffel Tower, which are copywritten in a way that is hard to navigate. Creative Commons has legal experts on staff who can guide on those matters. There is also the trickiness of people taking photos of something that is under copyright and making their photo of it openly licensed, for example photographing a cartoon or a poem in a book and passing it around the internet without thinking through what it means to have openly licensed this thing.

If this were 10 years ago, tagging and metadata would probably be manually created and maybe manually translated, but some of the image and vision APIs and cloud APIs with machine learning are kind of incredible. The plan is a first pass for tagging or identification from whatever a Google Vision or Azure Vision API can give back, put that in and use it to inform search, then use search interactions, what people search for and what they choose, to drive ranking, and then make it all wiki-like and editable, so if it thinks something is a hot dog that is not a hot dog, someone could go in there and tag that. For things like translation, they expect to do it completely driven by the algorithms and APIs, because tags of one or two words are well suited to machine translation, almost like a dictionary translation rather than poetry. He has not decided the right approach, whether to keep English as the canonical backend language and translate searches on the fly with a match to English, or store metadata for every language they operate in. The first is more efficient but he does not love English as a necessary canonical language. Creative Commons has run into words that translate directly but mean something different, for instance the word for elephant in Icelandic can mean either an elephant or a type of bird, which gives irrelevant search results without context. Matt mentions a website that is a dictionary of words that exist only in a single language, mostly cultural concepts rather than physical objects, including a Brazilian Portuguese one meaning the longing for something that is not gone yet but you know will be.

Creative Commons ran a pass through AWS Rekognition and it was fantastic, doing what human hours cannot compete with in a short amount of time. Two things are warranted: human QA just for good measure, to make sure the images are represented in a way you are comfortable with, and a wiki-like editing experience similar to the wiki meets blogs idea, where you interact with the creator of the image to help them describe it better and give them the power of approving those descriptors. They may have something specific and contextual to represent in that description that could be lost in a rapid or automated pass. For people who do not create but like to contribute, the Wikipedia model shows the power of heavy contributors alongside people who just add a link or translate something.

He had a GeoCities site and thinks he still pays Yahoo to redirect his old GeoCities site to a domain, because he likes the idea of things not breaking, and he should check if that still works. He also misses blog rolls, where if you had a website you put in your sidebar or a subpage a list of the bloggers that you followed, like a follower list, and you would join a blog roll like being part of a group. WordPress built something in, a very random feature that might even still be in the code somewhere, that would reorder that list based on who had most recently updated, so you could see people they like that recently updated and surf and explore in a web-like fashion. He wants to bring back the idea of a recommended list, because just things you follow is not quite the same: he follows people he disagrees with or would not recommend, because he likes lots of different viewpoints in his personal stream. The related feature he remembers is a link circle or blog circle, where about 10 humans with a shared interest all link to each other and you would see them along the bottom of the blog. WordPress still has pingbacks, which let you know when someone has mentioned or linked you on the web, but referrals are much less of a data source for privacy reasons, which has pluses but also minuses, because you do not know if someone blogged about you or put you in their blogroll. Linking to someone was a very exciting way to start a conversation or start paying attention to each other online. He calls the old identity model pseudonymous rather than anonymous, because there was definitely consistent identity.

The technical underpinnings of the web itself: plain old HTML and CSS and DNS and domains and websites are still the creative heart of the internet, no matter how hard companies try. Although mobile apps have taken some of that away, companies and people are now grappling with the Faustian bargain they made to be part of app stores and choose that as their only method of distribution, versus the freedom of customers coming to you through the web. People also spend much of their time inside Facebook or Twitter in a web view, reading a link to a web page or clicking through, and pretty much all commerce is still happening through web views or the web.

Gutenberg is the new editor, taking the idea and what he calls the beauty of HTML and making it more accessible to a layperson who does not want to learn how to code. Without using any code, they can create rich layouts, effects, widgets or functionality, whether that is a YouTube embed, a Google Map, or a complex layout with images and pictures. One feature coming at the time that is enabled by modern CSS is what they call duotone, which allows you to take any image and using CSS transforms turn it into a stylized two-color one, like a black and white photo where the black and white might be red and green. With a few clicks this is going to be in every WordPress user's hands. They build canvases and paintbrushes and paint, and the users create the masterworks with it, so whenever they find a new color of paint he gets giddy with excitement. They are about five years into Gutenberg, and he expects it to be the foundation of WordPress for another 10 or 20 years to come.

He recommends Krisp.ai, spelled K-R-I-S-P dot A-I, which he has been enjoying with everyone on Zoom and on calls all day. It is basically a $50 or $60 utility that turns any microphone into a studio where it will eliminate room echo, eliminate background noise, dogs barking, clapping and screaming. He says this is going to be built into the next versions of macOS and iOS, but in the meantime, for the next few months it is worth getting if you find yourself on a lot of calls in uncertain environments, whether that is the lawnmower going by or whatever is in your household. He finds that really high-fidelity audio gives you so much of the value of being in person, with a real-time conversation that sounds really good on headphones.

Automattic has been distributed from the word go. The first hire was in Cork, Ireland, so even the very first employee at the company was pretty far away. In the early days they did not do much audio or video, so they did not really notice the pluses or minuses of those mediums. They were primarily text-based interaction and communication, first probably AOL, then IRC, and then more recently Slack, and moving to something like Matrix, which he describes as a very cool program, like an open source Slack, and encourages people to check it out. Text has been the same as it was 20 years ago, while audio and video capabilities, particularly video, have evolved tremendously. The plan is to switch to Matrix, though everything is not finalized yet, because they try wherever possible to patronize open source solutions, and when they get as good as the proprietary alternatives, it is time to switch over. He also notes that people like working from home but forget that their home comes to their work, whether that is family, kids, dogs, lawnmowers or workers, though he likes it because people are literally bringing their whole selves to work.

It used to be that there were only a few good places for barbecue, and Houston, where he grew up, had almost no good barbecue. Now the top pit masters, including the guy who runs Franklin's, Aaron Franklin, started sharing their secrets and talking about their process, essentially open sourcing their methods for creating amazing barbecue, so there is a ton of great barbecue in Texas now and the average quality of barbecue throughout the world has risen. He has had amazing barbecue in Paris and Seoul, South Korea, and Texas barbecue. If you are in Austin, Franklin's is amazing and Terry Black's is really fantastic, where he was just last week. In Houston, his go-to would be Killen's if it is a little bit outside of town, which is maybe the best, and inside of town he likes going to Pit Room. Even the worst barbecue is pretty darn good now, and even gas station barbecue in Texas is pretty solid on a national scale.

He is photomatt on Tumblr, Instagram and Twitter, spelled P-H-O-T-O-M-A-T-T, and his blog is at M-A dot T-T, where he invites people to come and leave a comment. The episode closes with thanks to the musician Broke for Free, whose track 'Daybird' plays at the beginning and end of the episode, available under the Creative Commons Attribution License, meaning it is free for anyone to use, at the Free Music Archive, free music archive.org.