The long threads

That’s funny. I had driven I-94 many, many times. Growing up in North Dakota, I would drive it from Minneapolis to Minot. Any time I was coming back on I-94 East, I would see this sign. It said 3, 33, 99. To me it was just obvious that you should go up to the sign and put in a times sign and an equal sign, because here was essentially a math equation just sitting right there. I saw that sign dozens of times and thought, that’s so funny. Then I wondered how many other signs there were like that, with that kind of math equation in them. I thought, well, I should go look for some. And I found some. It wasn’t easy. They’re not that common. But I had a lot of fun doing it.

Then I think I was talking to friends, I don’t even recall, and somehow the idea of making it into a website came up. I wondered if other people would find it interesting. That’s when I created a wiki out of it and got other people going on it. But it was that original sign. To me it just seemed obvious that it was math, and I thought, huh, I wonder how many other signs there are like this.

It’s funny, the very first version is almost one of those stories about a rabbit hole I went down and took a wrong turn. I gamified it and created a system of points and scoreboards, and we had all these players and a leaderboard. It was interesting, but I took it the wrong direction by making it overly complicated, which took a lot of the fun out of it. It turned into something just not as exciting anymore. What put it away was a decision I made to simplify and rationalize the projects I was working on. It got swept up in that, not as a specific choice, but under the umbrella of having fewer things to tend to and work on.

realize the vision.

It’s entirely different because of AI. Without it I could not create the experience I always envisioned. Now, with the help of Claude and AI, I felt I could realize the mobile vision I always had in mind for this project.

Wait, before we go on, did you get my full answer to that previous one? I talked for quite some length.

I did. The second version was a direct response to a three-day road trip we made to Arizona. Tammy made these envelopes of activities to keep the kids busy. One of them was to play Road Sign Math, with a prize if they could find signs. I loved it, because it brought this thing I had built back into our family. The kids were playing it with their phones, taking pictures of signs and figuring out the math. It brought me back to the heart of the project, which is this fun activity, not all the rules and all that other garbage. So I decided to bring it back as a blog, keep it simple, and focus on having fun.

Oh boy, this is a big one. I love the open web. The web was born when I was in my 20s in college, and the delight of exploring it has always been something I truly enjoy. I still remember the very first version of Mosaic I ever used. It was unique in that every web server in the world was pre-populated in the bookmarks. That’s how early I got to see the web. So bookmarking has just been something I’ve done forever.

I’ve kept that database for many, many years. It became the genesis of the Weekly Thing: I wanted a way to publish the bookmarks I had found interesting that week, and that was the backbone of the newsletter. Having that published as a database has always been a wish of mine. I tried various versions, and none really landed as the thing I envisioned. I think the current version is what I really hoped it could be.

The idea came from a conversation with my friend David Hussman. We were talking about software versions, about having a major version, a minor version, and every day, and we were riffing on it and both really jamming on what a fun concept it was. Hence the blog post. I made it into a shortcut. I’ve done a lot with it over the years, but what I always wanted was something that brought it into my daily activity, so I’ve had a shortcut for years that shows it to me.

I think there’s a similar concept, by the way, with finitude and the 4,000 weeks idea. The pace of time and the finitude of life all connect together. To me, Your Version Number is a great way to bring that into daily life in a way that’s fun and interesting. Attaching release notes to it just seemed obvious. The software metaphor extends naturally, and it makes total sense to attach the things that happen, the release notes of every day, to Your Version Number.

I distinctly remember. We were at Sun Street Breads, having a conversation over breakfast. He and I used to meet pretty regularly, often over breakfast or a coffee, and talk about big ideas. I really miss him. He was a good friend, and gone too early.

To be clear, the Workshop never actually generated an issue of the Weekly Thing. I would say the Weekly Thing has had four eras. The first was when it launched on TinyLetter, the second on MailChimp, the third on Buttondown, and the fourth is still on Buttondown, but now with WT Builder and me taking over the archive and hosting it myself. The fourth is filled with a lot: WT Builder, taking over the archive, creating a generated podcast to go with it. There’s a huge expansion in this one. The first three are really all about me learning the content cycle and figuring out the right platform for sending the email.

TinyLetter was great. It was aligned well to my indie web ethos and to how I thought about hosting it. But it lacked core functionality. Most obviously, I couldn’t schedule the email to send in the future, and I typically wanted it sent on Saturdays at 7 a.m. It also had no API to send through. So I moved to MailChimp pretty quickly, I think after about a year, because it was far more powerful. I could automate sending the newsletter with the API, and I could schedule it. But MailChimp was a poor fit, because it was really heavy on HTML. It really is a marketing tool, not a newsletter tool. Once my content became HTML, I couldn’t really edit it. And in MailChimp I was sending a campaign. I’m not sending a campaign, and I have a strong aversion to advertising and all these things. I never felt at home there.

When Buttondown came on the scene, it seemed like an obvious and amazing platform. It was designed for newsletters. It was designed to allow individuals to send. I loved that it used Markdown, because then my authoring and automation environments could all be Markdown. So I created this complex sequence of shortcuts to generate the Markdown, and it all worked really well.

The biggest issue was that it was a one-way pipe. I could create content, but if I changed anything further down the line, it was really difficult to make that change back in the source material. The other thing was that it was not at all what you see is what you get. I built it in an entirely different environment from the one it was read in, which meant I could end up creating something that didn’t look right. And it was very rigid. I couldn’t easily change things on a single issue. All of that is why I ended up creating WT Builder, again enabled by the fact that I could build it with agentic technology. Otherwise it would have been incredibly expensive and far too time consuming to create an application only I would ever use.

It actually begins further back than you’re thinking, because it wasn’t just a data project. Elixir started as a Discord bot, and the intention was for me to learn about agents. What I wanted to figure out was: could an agent help me run and manage a Clash Royale clan? That was the premise. I realized that Clash Royale has an API that gives you read-only access to what’s happening in the game. That was a key unlock. Then I realized Discord would be an interesting control surface where the clan could congregate. So that’s where it all started, with a Discord bot written in Python using the Clash Royale API to help me and my fellow clan leaders run the clan.

The data project that became Elixir is essentially looking back at everything we built in the bot, realizing we could decompose it into individual capabilities that each worked better, and flipping the whole stack inside out: a data platform that builds product capability and then enables LLM agent capability on top.

I loved POAPs for a couple of reasons. One, I wanted to find ways to explore and use crypto technology to create things, and POAPs were a completely non-economic use case of crypto. I found it challenging that almost everything in the crypto ecosystem was fundamentally about putting a price on something and buying and selling it. With POAPs there was no price, no buying and selling, no economic activity at all. It was simply a protocol that served a purpose, and it used blockchain in an interesting way. It allowed me to assert that I did a thing without relying on any company or other independent arbiter to host and own that data. Instead, that data lived on the blockchain, which really delivered the promise of blockchain technology.

POAP informed my thinking about crypto in so many powerful ways. I became known in the POAP ecosystem. The founder of POAP and I have exchanged emails. I created many POAPs. I onboarded people to crypto using POAPs. It became something I truly cherished.

I’ve also always had a collector mentality. As a kid, I loved collecting things. I still remember collecting every single M*A*S*H figurine, for example. POAPs were a way of collecting, and the 612 POAP Challenge was the culmination of that: a game that ran on the POAP ecosystem, that I could pull people into and play in the real world.

The reality is, POAP was kind of doomed, I think, in part because crypto has never answered the onboarding problem. And the reputation of crypto became very sullied by all the fraud and theft there. I think it’s very difficult for any idea, no matter how good, to escape the shadow of the crypto industry right now.

Very simply, Semantic MediaWiki. I loved what you could build by bringing data and text together in a way that let you play with surfaces like they were clay. All of those sites were fundamentally built out of Semantic MediaWiki. You could connect the dots, by the way, back to AI and the fact that I’m just not a very good coder. All of those could have been built with software, but I wasn’t able to build them that way, and doing it in software would have meant one or two people having to do everything. Being able to create these surfaces dynamically at runtime and iterate on them very fast was super fun.

Ultimately, the MediaWiki ecosystem is not powerful enough to run these kinds of applications. It was too janky. If you think about how WikiApiary worked, hack is the best term you could apply to it. But I still think those concepts are super fun and a great, enjoyable thing to build. The semantic web and Semantic MediaWiki were a great idea that never really flourished the way it could have. I think microformats are probably the biggest real thing we still have on the web from that era. What ultimately made me walk away was that you could create these solutions, but they all fell under their own weight because of the complexities and challenges of MediaWiki underneath.

Garrick and I had a long run where we built a number of projects together. It coincided with when I was exploring Semantic MediaWiki, and we were really digging into Kubb. He and I and Jim Bernard played Kubb together a ton, and we brainstormed a lot around things like the Planet Kubb notation system and score sheets. Garrick is similarly passionate about technology, and about thinking broadly and exploring concepts. He’s a builder. He and I enjoyed building things together for many years and had a lot of fun doing it.

Let’s call it there for today.