Once I'd distilled my own taste into a toolset — I call it Wanxing — I kept building on top of it and eventually put together an MVP. Then I realized that nearly every idea and feature in it overlapped with Claude Design. There wasn't much point in continuing, so I dropped it.
A couple of small things did come out of it, though, both answering a real need.
Wenxin Editor
I've been publishing to WeChat Official Accounts recently, and Markdown doesn't paste cleanly into their editor. The third-party editors that exist for the platform didn't work well for me either, so I vibe-coded one that follows Wenxin's design guidelines.
🔗 Wenxin Editor
Fixit
Claude Design's annotation feature was worth borrowing, so I pulled it out into a standalone Chrome extension. When I'm building pages on top of Wanxing now, I can mark things up right in the browser with Fixit and turn the notes into a work order for the AI — "work order" is a strange phrase to find myself using, honestly.
Once the core features are mature and stable it'll go up on the Chrome Web Store. For now it loads locally in developer mode.
🔗 GitHub - Fixit
Afterword
Claude released Fable 5, and the pace of all this is genuinely hard to keep up with. More often than not, a flash of what feels like a really good idea turns out to be something someone has already built most of. Given how efficiently AI now compiles intent, even the strongest product managers aren't facing a mere step-change in leverage — it's erasure, in the time it takes to snap your fingers.
Lately I've been trying to work out a methodology for building AI applications, maybe something closer to a philosophy, but it hasn't taken shape yet.
So. That's where things stand.