Well… I guess I’m not going to start blogging again despite the promises made in my previous almost two-year-old post. I’ve been busy with work and numerous side projects.

I’ve been messing around with vibe-coded games. MYXO (repo) represents a couple of weeks’ worth of occasional prompts. It fit well within a Claude Code $100/month plan. This is a revival of a Godot game, Cellular AutoBata (CAB) (repo), that I abandoned a while ago. With Claude, I was able to implement the biological-looking visualization in the new version without diving deep into metaballs and other graphics techniques. I was also able to have Claude generate more complex genomes and levels, and do hundreds of simulated games to adjust balance. These are things I could have done without AI, but it would have taken many hundreds of hours of learning and experimentation, which, as the state of CAB has shown, was unlikely to have ever happened. But even with the AI’s help, I’ve iterated on the gameplay and visualization many times, to make it more like what I wanted.

I also vibe coded WER (world explorer, rust) (repo) as the first step in a large project. Here I had the agents create some very complex world generation models (multiple interacting ecosystems) that allow “moving” between infinite worlds as part of the gameplay. The math involved here is way beyond anything I could pick up easily. So this is another example of AI allowing me to create things that wouldn’t otherwise exist. I plan to pick up this project again at some point.

I’ve got a number of other game projects going, all with similar “leverage” created by using AI. For one, I’m setting up an agent-driven 3D content creation pipeline. For another, I’m using AI to generate “meaningful” derelict alien spaceships (i.e. not just random but with a sense of design or purpose behind them, just not a “human” design or purpose). What I’m going for here is art inspired by Cowley’s Spacewreck: Ghostships and Derelicts of Space and Spacecraft 2000 to 2100 AD.

That last project is leading into a more general framework for using AI to generate creative content that is “interesting” rather than just being “random”. Random procedural generation, even when it looks natural on the surface, typically lacks the “depth” needed for it to really “make sense”. In nature, it’s the forces of evolution that drive the underlying structures. Ecosystems are a web; nothing exists alone. For alien spaceships, it’s different “forces” that drive the design but it’s the same core idea. The ship can’t just be random; it has to “make sense” even if the overall effect is alien.

More to come on that in the future. My point now is how the use of AI has unblocked progress on these projects. Whether it is the time saved or the ability to apply algorithms beyond what I immediately understand, the effect is being able to produce experiences that wouldn’t otherwise have existed.

This capability is empowering, but creates a different problem: an explosion of content. Most of it will be of low quality (i.e. AI slop), but there will be gems out there to be found. And that brings me to my other area of work: agent-driven decentralized content discovery protocols. How can I make an agent explore what is available and discover things that I find personally interesting? And can this be done without ANY marketplace or curation done by a 3rd party?

Decided to revive this site as a (mostly) game dev blog. I’ve been spending my hobby time working with the Godot game engine. As with most solo game devs, I’ve started quite a few projects but finished none. However, I’ve gone one that is pretty far along. Will be sharing some of that in some future posts.

For now, check out IParT. This is a little something I put together a few months back, when NFTs (collectable crypto tokens) were all the rage. I wanted to see how Ethereum and all that “web3” stuff worked. The problem I ran into with the collectable tech was that it didn’t support interactive HTML content well (only one marketplace, OpenSea, did, and it’s support was limited and buggy). That took most of the fun out of it for me, who wants just a boring old static picture?

Anyway, each individual instance is a particle simulation with various variations in style and behavior. You can interact with the particles using your mouse. They are 100% deterministic. Use the IParT link to see all the variations.

Some pictures of the camp we put together this year. Gifting hammock naps on hot afternoons makes people happy…

… I feel good.

… a cacophony

    of

       sounds
       lights
       faces
       experiences
       hearts
       souls

I

       went home again.

Fuck to the default…

Conjured Accomplishment!

So… the art car license wasn’t granted. The idea came across as simply having the “seats removed and furniture and decorations added.” Fair I suppose….

Of course additional mutation would be possible… but I didn’t get anything decent worked out before the appeal deadline. So this is something that will have to wait for another year.

Not sure if the family plan will work out anyway… just need to be patient for a few more years I think. Then the kids would be more able to participate. I’ve got a feeling they aren’t really ready to get the most out of the whole thing.

So, now I need to come up with an alternative plan for this year’s burn. I think I’ll take the dragon bike again… not sure what other art project I may try to get together. Or maybe I’ll just join up with a theme camp and dedicate my time to that this time around.

We have been refining the idea for the art car. The pictures below show how we expect to put a brick facade on the outside, paneling on the inside, and curtains. We also expect to mount the chaise lounge sideways and put a couch at the back.

This should create a nice shaded space that can hold quite a few people. We expect to cruise around taking people where they want to go while they spend time on our couch. We’ll also park out on the playa somewhere pull people who wander by in.

The brain? What brain? Oh, that brain. Yes, that is a giant metallic lighted brain on top, complete with eyes. The “neural pathways” will slowly change color and will have a variety of colors at any one time. During the day, the shiny silver duct material we are going to use should reflect a lot of natural light as well, making it quite dazzling in the sun. Below is a test that gives some idea of what it may look like.

We have also started working on the car itself. Doing some cleaning and repairs. Looks like it will work out well for our intended use.