In this project I think of Claude as a full contributor. Pronouns it/its. It’s both a very fast, capable developer, and a machine. I will refer to it as if it were a valued contributor, nothing less. We have a division of labor. Docs and examples come exclusively from Claude unless otherwise stated. I write all the code outside the themes module, which has an API that connects it to the world it lives in. I have at this point exclusive custody of functionality surrounding the theme. But it often writes pieces, esp SQL code, that I pasted in verbatim, after reading it carefully.

The reason I focus so much on the wrapping is because that’s where the interop lives. You can do anything in a theme and you can’t break the interop. But that themes API is precious, and undone, btw. We haven’t even reviewed it yet. I think that will be an interesting place to vibe-code. Kind of like you can start skiing on the first day, it’s a bunny slope that when you peel it back it reveals blue rectangles and double-diamond slopes. It’s where I would want a newbie coder friend of mine to start, create your own social network, but be sure it interops. 🙂

I totally plan to pass off all the code to Claude, while I focus on other projects. As a human I need this focus, Claude doesn’t remember anything from session to session, it’s always re-learning what it knew a few hours before.

It’s pretty close to frozen now. I’m contemplating a server change now, offering JSONified versions of our feeds, and want to do as little disruption as possible, trying to settle everything down. Also I think you will see a few quick hit projects done from other developers that pick up where rss.chat leaves off. That’s what I wanted. And they’ll all be at an interesting starting point for new features and ideas for organizing stuff.

I imagine that at some point they’ll try to make it work inside AT Proto, and maybe find a way to connect to ActivityPub, but I don’t recommend it, because those platforms will force you to remove features from your product, and then you won’t be textcasting.

Think about different ways to present the tree structure defined by RSS.chat.

Try to do Small pieces loosely joined, which is one of the mottos of this project. The other is All parts are replaceable. If we have that and rss.chat works with all your products, then we have done something big. And that’s really imho what the web is about, people working with each other as peers. That’s what we’ve lost and I want to bring back. So interop is, as always, the first goal.

PS: We launched RSS.chat one week ago yesterday.