site banner

Tinker Tuesday for July 28th, 2026

This thread is for anyone working on personal projects to share their progress, and hold themselves somewhat accountable to a group of peers.

Post your project, your progress from last week, and what you hope to accomplish this week.

If you want to be pinged with a reminder asking about your project, let me know, and I'll harass you each week until you cancel the service.

1
Jump in the discussion.

No email address required.

I've successfully made a second, much higher quality mod for Rimworld. It adds 19 new guns, half of them railguns.

I really like railguns, okay?

But then I branched out and added cased telescoped ammo, ETC weaponry, 2 40k Bolters, a laser rifle (where's the rifling?), a plasma gun, a 25mm smart grenade launcher, and a 20mm infantry autocannon. Why? Because it's my mod and I do what I like.

I can already notice the massive leap in capabilities in less than a year of AI development. The LLMs of 8-9 months ago were inadequate for making a niche mod without an enormous amount of hand holding. The last time, I caved, and just did the XML work myself, and used AI images that I wasn't quite happy with.

This time? 5.6 Sol did pretty much everything I asked of it, with some mandatory manual testing to find and share errors and balance issues. It took me several days to implement the 4 weapons I'd made last time, in half the hours, I made 19, of significantly higher quality in all regards. I even managed to create brand new types of ammo, which I never managed to get the hang of last time. The art is far, far better. 🅱️ anilla Rimworld has art of the quality you'd expect at a special needs art therapy session, so it's not like it's hard to do better.

All I had to do was guide the overall design, use my knowledge of existing mods to direct the model to find out the poorly documented little extras or tacit knowledge that it wouldn't know, and make sure nothing was glaringly broken. Along the way, it even managed to troubleshoot and fix a 450 mod Rimworld map-gen issue, which the models of last year could not do.

It was fun, albeit tiring. I intend to add more guns, and now that I know it can handle more ambitious challenges, turrets, guided projectiles, and pretty much anything else I'm willing to burn time and tokens on.

Predictably, this caused drama on the modding discord. I was expecting it. I told the modder with a fetish for fetish art in the thumbnails of completely unrelated mods to go fuck himself, or thereabouts. I'm making free content, and it's good stuff. End of line.

I've also noticed Opus 5 has improved even since the last big jump in November. It's seriously impressive now: I give it a task and it does it, writes good tests and makes them pass, all on its own. I'm writing barely anything by hand, although still read its output and ask for refactors, because I'm still noticing subtle mistakes (maybe all spec issues, my fault, but still) and redundancies (definitely not spec).

My moral principle is that publishing AI content is good, as long as it's disclosed and you avoid places that explicitly disallow it. Don't be one of those assholes that pastes obvious LLM output saying everything is 100% human, just state upfront "I used AI because I wouldn't have made it otherwise".

I say nothing at all on the matter. The mod is exactly as advertised: it adds a bunch of guns to Rimworld. The guns function, and look good. It is immaterial to me how such a product comes into existence, as long as it does what it says on the tin. God knows I wouldn't have bothered to make anything of that scope if I had to do it all manually. I could, it's not impossible, and I've made one mod on my own, but something like this wouldn't happen.