The ones that ended

Most of those I don’t have much fondness for. But VeloTweets made me smile a lot, because I had forgotten about that project. It reminded me of the best of times for Twitter. When there were tons of professional cyclists using it, Twitter was just fun. It hadn’t become the gross platform I think it is now. It was a great, fun project I built with a couple of friends. It had a good run, and it made tons of sense as a way of aggregating things around a collection of users, which predated what a lot of Twitter clients ended up doing with lists. We were essentially the precursor of much of that.

Foreground and Filament are fun because they remind me of the MediaWiki era and of focusing on building something compelling around that. I think it’s cool that Foreground was eventually adopted by the Wikimedia collection. It’s now managed by Wikimedia and has become a supported project of that ecosystem, and a number of MediaWiki sites use it.

And I will always remember Flight Path, because it was one of my first agentic projects, vibe coding, if you will: pulling the POAP ecosystem together into a 3D visualization of activity around the globe. It really was a wow for me, what I was able to do and how AI could help me build solutions in ways I would never have been able to. Flight Path combined a lot of things that were not that hard to connect, but really difficult for me. And I was able to do it literally via prompt on my phone and have it up and running in a couple of hours. It was incredible.

TOSRedux really is about embracing the privacy values I have. The idea was that we should be able to crowdsource, among a lot of people, what terms of service and privacy policies actually mean. Combine that with a browser extension that puts in plain English what the website you’re about to engage with is asking to do with your data, and it would help people make a more informed decision about whether they want to be on that platform. My belief is that many people have no idea what the terms of service are saying and what they actually mean. Doing that on a wiki was interesting, but there’s friction all over the place, and ultimately you have this problem of: does anybody really care? I’ve become fairly cynical. While I believe privacy is really important, I don’t think the average user gives it much value, and I don’t think it would have gotten much adoption. Today I would look at that and say it’s easily done with AI. But I think you’re swimming upstream with a general public that doesn’t think anything is private anyway and has kind of given up, frankly. I refuse to give up when it comes to privacy.

The other one, Weekly Thing Studio, was interesting because I felt like I could use AI coding tools to maybe create anything, so I didn’t really know what I wanted. I framed it as the idea that a creative studio filled with agents working on projects alongside me could be really powerful. Ultimately, it never worked. I tried having a number of different agents working and collaborating, and I was never able to get that to produce anything of value. I abandoned it very quickly, because it always became overly complicated.

You could say the Studio is very similar to what Elixir Bot originally started as. Elixir Bot was an agent to run the whole clan, and then I flipped that completely inside out, building Elixir as software and databases with an agentic layer on top. With the Studio I tried to build an agent solution for everything, and ultimately realized that WT Builder and the weekly website were the better solutions, with a thin agentic layer on top, which WT Builder has.

These projects all came to a natural conclusion. Open Loop was a podcast Garrick and I did, and we decided it was time for it to come to a landing. The Charlie Brown Tree was a very finite project. We had it carved out of a dead tree at our cabin, and it’s an amazing piece of art. Jiggy Thing Puzzles was a funny one, because we realized how hard it is to make a jigsaw puzzle. It took a tremendous amount of effort to design, and then the costs of producing a puzzle at subscale are just crazy. We realized we were spending the same amount of money we were raising, because if you buy fewer than a thousand, the costs are way too high. So all of them had natural landing points, or it became obvious they weren’t worth pursuing.

It taught me two things. One, sending a bunch of data into that system was not going to give me the experience I wanted. Two, it gave me enough of a taste that there was something here, and that if I could build a real agent around it, that could be meaningful. The basic ability to copy and paste a ton of files or JSON into the GPT just wasn’t robust enough for what I really wanted.

Probably the Foreground MediaWiki skin, because it’s adopted by Wikimedia, and a lot of websites and MediaWiki installs run it. I know for sure it has the most GitHub stars of any project I host, by a long way. And generally, no, that doesn’t mean a lot to me.

Let me change that answer, though, because right now there are over 1,600 people who subscribe to the Weekly Thing, who read it and engage with it on a regular basis. So by far the largest audience I have now is the one that reads the Weekly Thing. It’s the thing the most people engage with, by a long way. Even against actual software projects, I think the Weekly Thing is by far number one.