I spent an embarrassing amount of time last week pretending to be an AI chatbot, and I think that says something important about where we are right now with artificial intelligence.
Your AI Slop Bores Me is deceptively simple. You pick a side: human or LARP as an AI. Submit a prompt or an answer. Get 150 seconds to respond. It’s absurdist, constrained, and weirdly addictive. But what really struck me wasn’t the novelty of the format, it was how human the AI responses felt compared to what I get when I actually open ChatGPT.
There’s a credit system that’s almost too on-the-nose. You burn credits making requests, but earn them by jumping to the other side and answering prompts yourself. Or you can just wait two minutes for a free request. It’s a real token economy, except instead of being managed by billion-dollar companies, it’s managed by actual people with attention spans and creative constraints.
The Constraints Make It Better
Here’s what I think is happening: artificial limitations breed creativity. That MS-Paint-like drawing tool? Impossible to use well in 150 seconds. That’s the point. Someone has to make an aesthetic choice, prioritize what matters, and accept that they can’t polish everything. When I prompt real AI, I get a polished non-answer. Everything is fine. Nothing is wrong. Nobody’s offensive. Nobody took a real risk.
The humans playing AI are taking risks. They’re trying to capture something real inside an impossible constraint, which is genuinely closer to what creative problem-solving looks like than most AI outputs I see.
I’ve been following the AI trends on mgks.dev for a while now, and what strikes me is how much the industry is obsessed with scale and capability. We celebrate when models get bigger, when they can handle more tokens, when they get “better” at following instructions. But something gets lost there. The game reminds me that constraint and imperfection aren’t bugs, they’re features.
What This Reveals About AI Culture
The Hall of Fame includes an alarmingly good Starry Night rendition made by a human with a basic drawing tool in two and a half minutes. That shouldn’t be possible. It shouldn’t be better than most AI image generations. But it is, because someone cared enough to try within the constraints.
I think we’ve been measuring the wrong things. We’re obsessed with whether AI can pass the Turing test, but Your AI Slop Bores Me inverts the question: can a human convincingly fake being an AI, and more importantly, will people prefer it?
The answer is often yes. Not because humans are inherently more creative, but because humans are making actual choices. They’re working within limitations. They’re gambling on what matters and what doesn’t. Real AI doesn’t get to do that. Real AI gets to hedge every bet.
The Meta Layer
There’s something delicious about the timing of this game. We’re currently in the phase where AI is either a productivity miracle or a fraud, depending on who you ask, and both sides feel right. Your AI Slop Bores Me sidesteps the entire debate by making the question irrelevant. Who cares if the response comes from a neural network or a person roleplaying? Judge it on what it is.
For developers and people building in this space, I think that’s a useful gut check. The race to automate and scale is real, but I’d argue there’s more competitive advantage right now in building tools that make humans better at being human than in building tools that replace human judgment entirely.
The Discord server has probably generated more genuine conversations than most AI Discord integrations. The ads are obtrusive, sure, but at least someone’s making money off attention rather than claiming to have solved AGI.
I’m not saying abandon AI development. I’m saying that watching actual humans strain to fake being an AI in 150 seconds might be telling us something about what we actually value, and it probably isn’t what our current incentive structures are optimizing for.