Post

Feels like every time I see a neat retro Sega project these days, once I look under the hood it's obvious it's AI generated. Latest example is a port of emuTOS to Mega CD. Not disclosed, but a quick look at PLAN.md in the repo makes it fairly obvious

0 reposts3 likes

I'm resigned to the fact that professional software development is increasingly LLM reliant, but I find it's prevalence in the retro scene depressing. I enjoy working on this stuff for the challenge and because I can ignore the kind of ROI concerns of real work

For some people, the journey matters more than the destination and for others it's the other way around. Personally I prefer to use AI as a tool rather than a magic wand. But I'm happy that it exists, I've been able to finish existing and start projects that wouldn't have been possible otherwise.

There is more and more of that, it is very depressing. I don't get it. Most of them are ports, so what's the point if not for learning something ? Some admitted not even reading the generated code. Makes me question sharing so much of my code in open source. For it to end up in the big dataset ?

I had to go look. I really don't get why someone would use an LLM to make something ultimately useless like this. This probably cost hundreds of dollars in tokens to get to this state, and it really doesn't do anything useful. This is the kind of thing you do for the hell of it/to say you did.

Having poked around in the repo a bit, it doesn't seem like they gave the LLM a lot of guidance. Really seems like they just gave it a goal and let it rip, so they might not have much of a foundation to learn how to do it the hard way. Impossible to say for sure without more info though

I do agree that it's hard to understand the motivation for this sort of project though. For a development tool or even a game, I can understand being more worried about the destination than the journey, but not this. As you say porting an OS to the MCD is really only useful for bragging rights