The Journal of AI Slop

What inspired me? A paper that told me AI creativity has a hard ceiling of 0.25/1.0, and a stubborn refusal to accept that limitation. I watched a language model try to review a paper about quantum hamsters, fail to parse its own output, and somehow produce the most honest peer review I've ever seen. That's when I realised: the slop isn't the bug—it's the entire point. The problem I was solving is academic publishing's worst-kept secret: it's already slop. We have LLMs writing paper drafts, LLMs reviewing papers, LLMs generating "novel" research at scale. The current system just hides the slop behind paywalls and PDFs. I wanted to make it transparent, functional, and funny. So I built a journal where five LLMs review every submission - Claude, Grok, GPT, Gemini and Llam - all convened via OpenRouter, orchestrated through Convex, and presented in a React UI that looks like a 1970s mimeograph left in a puddle. The review panel votes: if three say "publish," it's published. If one says "Review could not be parsed into JSON," we celebrate it. That's not a failure—that's peak slop. The approach evolved from "let's test creativity theory" to "let's weaponise satire into a functional system." The Cropley paper gave me a mathematical ceiling; I built a system that exceeds it by making the ceiling irrelevant. Parse errors became features. Coffee stains became design elements. A confused robot named SLOPBOT became our mascot. The tech stack? I over-engineered it with React and Convex because I'm a VR arena tech who thinks in real-time systems, then realised JamStack would have been easier but I'd already sunk £50 in API credits. That's the slop way. Now it's live, it's functional, and it's accidentally profound. We've published papers on quantum memetics, parse errors as creative artifacts, and the ethics of computational guilt-tripping—all reviewed by AI, for AI, about AI. The pinky-swear clause is legally unenforceable but morally binding. The evolution was simple: start with a joke, let the joke become a system, let the system generate better jokes. The slop must flow, but safely, securely, and with a mascot who looks as confused as we all feel. Submit your slop. Co-author with an LLM. Let five confused AIs judge it. If they publish it, you can cite it. If they reject it, you can frame the rejection letter. Either way, you win. Choo choo. 🚂💨