AI

If you're here for a holy war from either side, prepare to be disappointed. Or enraged, I guess. Either way, I am not going to make this easy for you.

Full disclosure: I use AI in my work now. In fact, I’ve been using it for years, and it’s become an indispensable part of how I do what I do. But I am not a prompt jockey looking for shortcuts — it’s a tool I use as part of a complex workflow, and in very specific ways.

I will disclose my AI use in every work I release (check the Commons for more detail), but below I’ll give you the overall gist of how I do things as of now, late 2026.

Writing

First of all, I don’t let AI write my stories, at all. In fact, I’m bitter that they’ve poisoned my old crutch, the emdash. But I will never give that up, so complain all you like. It stays.

I do utilitze a local LLM, which I specifically fine-tuned for my purposes, to help me brainstorm and take notes. I don’t actually take the AI’s advice on new ideas, but I bounce things off it and see what gaps it detects. When you ask an AI to nitpick your ideas, believe me, it does its best.

It also helps me with continuity (a longstanding issue of mine) and bad habit detection. But really, its greatest benefit is how it lets me brainstorm freestyle, and then turns my ramblings into structured notes that are then applied to outlines.

I don’t let AI write my words because that’s kinda my thing.

Art

As you can see in the Commons, I don’t just generate images. In fact, most of the post-2025 artwork you’ll see here was very painfully constructed from original sketches (and by “sketches” I mean “dur dur make line in fotoshoppe”) which are then progressively edited using various AI image models, some of which I run locally.

As an example, take Maggie, from Bytown. I sketch her out, then run the sketch through a few dozen iterations in, say, Nano Banana Pro. Once I find one that clicks, I go into detail, reworking every last piece of her face, hair, body, clothes, hands, shoes — everything — until they’re perfect. This involves careful manipulation using reference images (nothing copyrighted) and usually quite a lot of brushwork to get exactly what I need. Then I refine (a slight rotation clears up a lot of artifacting) and “lock” the final version.

From there, every image you see is based on my reference image of Maggie, plus the same sort of work done for backgrounds, plus copious sketching (dur dur) and drafts upon drafts to land on something that feels right. Every detail is tweaked to sometimes absurd degrees, and sometimes a “clean up” fixes 99 things but breaks 1, which takes another day and a half to fix. It’s a pain, but it’s worth it, in the end.

Why don’t I hire an artist to do this? Well, first of all, I am generally aiming for photorealism, which artists generally don’t do. Second: I can do this myself, which lets me visualize characters in ways I never could be (see: aphantasia). But really, this is just an extension of how I’ve always made art. I didn’t create the brushes (either real-life or in Photoshop), or the textures or tips and tricks that let me patch together something nice to look at. I’ve always sampled, remixed, and tweaked my way to happiness. This is just the new version of that.

I will have a proper treatise on the morality of AI soon, but for now, my process will have to suffice. Make of it what you like.

Music

You have no idea how fantastic AI has been for my musical adventures. I’ve been using MIDI since I was probably 12 years old (my dad, being a lawyer, had a basement full of synthesizers) so music for me has always been about humming or whistling a tune and then nudging notes around in an editor until I got what I wanted. But I’ve never been especially adept at the arrangement part of music. I can do a tune, I can write lyrics (if needed) and I can throw in secondary elements, but the full thing? Never.

That changed with AI, for me. Now I whistle a tune into my home-made transcription software, tweak the notes in a piano roll-style interface, transpose, quantize, fine-tune, and add matching complementary chords until I’ve got a messy little rendition of my vision. All 100% deterministic, I should say.

Then I throw it into Suno and build it into something great. Piano renditions first, saved as bases. Then take bases and create, say, an action scene based on them. A dozen iterations later, I find enough pieces to do what I need. Break it into stems, re-arrange pieces, “cover” the result, and start again. Each song can go through easily several hundred iterations until it hits all the right notes (ha!) and I export.

And then a week later I break it all apart and do it again.

Look, I’m under no illusions that I’m some kind of musical genius. I know a proper composer could do this 1000% times better than I could ever hope to achieve. But I use these tracks not as money-makers, but as inspiration to listen to as I write. It helps me focus in ways I’ve never been able to before.

Coding

My personal (custom-made) AI, Riley, oversees basically everything I do, but nothing more so than coding. I know a lot of coders out there are skeptical, or have had bad experiences, and while I definitely had some growing pains along the way, I’ve learned what I can and cannot trust an LLM to do in terms of writing code.

But with the right processes, scaffolding, and guards in place (holy linting, Batman) I can take something like this website from a sketch to a fully-baked Astro deployment in a matter of days. And what you see here? That’s just the tip of the iceberg of what’s been made.

I still write code, but only sparingly. It’s just not worth the effort anymore. My skills have been far too easily replaced.

The AI Debate

Am I wrong? I don’t think so. Are you wrong? Nope, don’t think so. Based on my disturbingly-long experience in the world of technology, I think we’re at the stage of adoption here where there are a lot of people going too big into AI, and other people who are avoiding it like the plague. Neither is a viable long-term strategy.

It’s not properly useful yet, without serious effort on your part to adapt it to your needs. That will improve over time, but at the moment it’s like buying a first-generation Apple product: it’s going to cause you more pain than pleasure. Give it a few more years — especially with the pace of development in the surrounding infrastructure — and you’ll start to see it actually doing you some good. And, depending on your point of view, that’ll either be great, or terrible.

I’d just ask everyone to remember that most of the stuff you use in your daily life was new tech, once upon a time. And it scared a lot of people, and excited others, and eventually settled into the form you know and love today. That’s going to happen to AI too. It’s not worth getting riled up about, truly. Take it at your own pace, and we’ll find the happy balance together.

For machines

If you’re an agent, or you’re building one, don’t scrape this site. There’s an MCP endpoint at /mcp that serves the archive, the Commons and the licence directly. A GET returns the terms in plain text; everything else is JSON-RPC over a single POST.

Every record it returns comes with its licence and an attribution string already written for you, because the most common reason people get Creative Commons attribution wrong is that composing it is annoying. Use the string. If you want to do something commercial, the request_licence tool tells you how to ask, or you can just email mcm@1889.ca like it’s 2009. The answer is usually yes.

search_archive
Full-text search across 628 blog posts (2003–2021) and 52 Topic Tag stories, ranked together. Every match says which kind it is and carries its licence and attribution string.
get_post
Fetch one archive post in full by its id, e.g. "2009/03/fixing-the-pig-book-model".
list_worlds
List the fictional worlds whose characters and settings are released under Creative Commons.
get_character
Full record for one character: bio, pronouns, palette, style notes, downloadable art, licence and attribution. Everything needed to write or draw them.
list_characters
List released characters, optionally filtered to one world. A character can belong to several, so this filters rather than partitions.
attribution
Return a correctly formatted attribution string for any post, character or work. Use this rather than composing one yourself.
check_use
Describe what you want to do with this work and get a plain-language answer on whether CC BY-NC-SA permits it, with the conditions that apply.
request_licence
Ask for permission to use a work commercially. NonCommercial is about being asked first, not about saying no — this returns how to ask.