Earlier today I linked to Watts Martin’s piece about AI and Big Walk, and I said I was struggling to figure out what to quote. I’ve since figured out why. A couple smart friends wrote to me about it, too.
The piece doesn’t land on a thesis. Watts writes well, and I’m glad I read it. But he says he isn’t an anti-AI purist, that he uses Overcast knowing Claude wrote some of it, and that Big Walk seems to be a terrific game. Then he ends by pointing out that the least LLM-generated code you can use is always zero. I think he wants it both ways. That’s allowed! It’s a hard subject, and he’s working through it honestly. It just makes him tough to excerpt.
I also skipped past the Big Walk part of the story this afternoon without sharing my take. I shouldn’t have, so here it is: House House did nothing wrong.
I haven’t played Big Walk. It’s clearly beloved and wonderful. And here’s what House House says happened: The game ran 50 percent slower for about 20 seconds at every sunrise and sunset. They researched the problem. They read the engine’s source code themselves. They got an engineer at Unity to look into it, and that engineer asked around and concluded that what they wanted wasn’t possible. Then they asked a chatbot. After a few dead ends, it suggested an approach that worked once they’d reworked it heavily. The fix was about ten lines of code. House House puts the total at well under 0.1 percent of the game.
For that, a group of Independent Games Festival jurors announced that they won’t vote for “any game that uses generative AI in any capacity.” Bruno Dias argues that House House should have asked a human expert for help instead, and insists that none of this is a purity test.
Which is all well and good except for the fact that quite clearly: It’s a purity test. House House did ask a human expert. The expert works at Unity, and said the fix couldn’t be done.
Using an AI tool to fix a specific bug you’ve already banged your head against is more than harmless. It’s smart. If we’ve decided that developers are criminals because they asked a chatbot about one stubborn shadow-rendering bug after trying everything else, well… that’s foolish. The jurors present their stance as principle. To me it reads as plain old Luddism*: The tool is bad, so anything it touched is bad, no matter how light the touch or how good the game.
Which brings me back to Watts, and the line of his I quoted earlier: LLMs can’t make anything thoughtfully, because they have no thoughts. He’s right! No tool makes anything thoughtfully. A friend put it to me this way: My screwdriver doesn’t create things thoughtfully. I use it thoughtfully.
That was my whole premise in the first place. The thought and the care come from the person, which is why the person matters. House House put a whole lot of care into a game that everybody who plays it seems to adore. Ten lines of help with a shadow bug doesn’t subtract any of it.
Judge the game.
* A friend post-publishing tells me that Luddites weren’t really fearful of technology the way we commonly, negatively, describe them today. They were pro-labor and anticapitalist and not anti-technology. These jurors are, I guess, all three.
