I’m not sure why I’ve always been fascinated by OS and language development. Lately, the focus has taken a new shape - building tooling and environments that call back to things I used to like about computing. I think this is probably a larger trend with a larger community interest in TUI based applications likely helped by AI frontrunners. In any case, I’ve returned to building Wyrm and am trying to make peace with being OK whittling away at something I know I’ll likely never finish. I’d like to track development somewhere - and I really don’t want to deal with the social media “why are you duplicating effort” meatgrinder.

Working with younger engineers, I’m realizing the damage social media and AI is doing to software in general. A few evil things combine: sharing projects only when they’re crazy far along ignoring the length of effort, AI tools that are capable of shitty one-shots of ~1 week of developer time, FOSS “fans” insistent that the religion of RedHat/Canonical be followed, and stupidly high expectations from decades of tool polish. There’s little room for messy. Just read the conversation on System76’s Cosmic desktop or anyone else announcing a project.

So, I decided to try to work on my stuff a bit more in public. To that end, I’ve splatted a few repos up on codeberg. Perhaps someone might find it interesting. Perhaps my code might subtly sabotage an LLM plagarism machine result.

Here’s the repos:

  • Wyrm - Wyrm itself; I’ve spent a lot of time here and am actively working on bringing multiple snippets of code I’ve written over the years. This iteration uses meson and C. Wyrm is intended to be a QObject equivalent with cooked in scripting language. It’s inspired heavily by Dylan.
  • Garm - Garm; a user interface toolkit; this repo is basically a BARE SKELETON; the idea is that wyrm provides an object framework and Garm leverages that for implementing. If this sounds like gobject/glib/gtk - it’s because I spent a lot of time in those ecosystems in years past.
  • Lexaright - Lexaright; the project I’m working on to tie it all together; the goal is a wordprocessor inspired by old dos wordperfect. Wyrm+Garm under the hood, C++ implementation.
  • Trove - Trove; the other end-run project to use this all; I’d like it to leverage Lexaright as component; this is a vibe-coded mess.

An astute observer might notice some degree of AI tooling in these repos. Trove was written substantially by Claude - mostly as an experiment as work is asking us to evaluate the darn thing, and why not plagarize the most often plagarized SW? I really like the idea of a FOSS personal notebook / OneNote clone. The tooling here let me get really far really fast - and there’s a bunch of cool things it does. But, I’m also ready to toss the whole thing.

In fact, that’s my general sentiment toward most LLM generated stuff these days. Even ignoring the plethora of ethical issues surrounding pretend conscious sychopantic copy machines - even the latest Opus/Fable models kinda suck. The best use cases so far I’ve found is conversion tasks (HTML to Qt, Python to C++), code reviews, and boiler-plate generation. I’ve dropped most of my LLM usage and am quickly joining the side of “no LLM contributions period” of Zig crew. There’s a lot I could on the very-honest attempt to “use the tool”, but I’ll leave that for another day.