Standup the first!

So, I’m out here looking for work, and I really miss my standups with my old boss. So this is me writing what I did, what I’m working on, and some rambling.

Ack! It broke!

I pushed out my Color layers blog post recently. It’s a little web tool to play with indexed colors in images and my thoughts on some trauma—and how doing some art, even if I didn’t think much of what I was making, helped. Then I immediately realized I did a terrible job with cross-browser testing. So in the spirit of the article, I’m leaving the article up but pulled down the LinkedIn post so as not to be totally embarrassed by it.

What I did last week (or recently enough to talk about)

Project starter

I got my project starter out. I might write a blog post about it, but the gist is this: it’s a GitHub template with GitHub Actions for versioning based on conventional commits, with Addy Osmani’s Dev Skills preloaded, and a new LLM workflow I’m refining based on code changes as narrative. Basically, the idea is, “Hey, I have this huge thing I want to build. Turn it into a story with chapters.”

Using this approach, I think my code quality has gone up, and I understand more of what’s happening. Breaking up code changes is a good idea anyway, but this is a direct response to Andrew Murphy’s article: If you thought the speed of writing code was your problem - you have bigger problems. It’s a great article. The thing I pulled out of it is: we’re generating code faster than people can take responsibility for it. So I’m trying to come up with my fix for that.

My buddy pointed me to Phabricator, which also does some of this, so I’m looking into it and seeing what I can pull over.

Astro/Sanity starter

I’ve wanted a ready-to-go template for building Astro projects with Sanity. The backbone of it is to get an array of content objects and map those to components. It’s built on top of my project starter above, so it was a good way to test out my new LLM workflow and my conventional commits.

It’s also got some CLI tools set up to build out Sanity documents that are already wired up to Astro page templates and previews.

Up next

Automated/browser-based accessibility testing

I’ve got some accessibility work coming up, so I’m starting work on my point of view on automated accessibility tooling. I’ll probably have a first pass by next week. Since the only JS that older CMSes can run is in the browser (looking at you, WordPress and Drupal), I’m building it to run in the browser first, which should open it up to work as a plugin for popular CMSes, as well as web extensions or bookmarklets. That should support shifting accessibility left.

Radio House OS

“OS” is a little cocky, but basically every person I know who’s working with LLMs has some sort of “the model eats all my notes and tells me important stuff when I need it” setup. This is my version of that. It’s basically a file editor/viewer where I’m pulling tasks that don’t need an LLM out of LLM handling, because I think LLMs work best when they work least.

The first thing I’m pulling out is todo/task handling with deadlines. There’s now a regex that runs on anything we’re keeping in the context and uses that to generate a list of todos without any LLMs involved. It might return false positives, but so far I haven’t encountered any.