[01] The Diagnosis
AI slop—or the more technical term "AI slopification"— is inevitable, given the current number of companies that are determined to adopt more and more users everyday. The only thing we humans can do, we're afraid, is either to live with it — which we, ourselves, have found such a difficult thing to do — or we can make it more accommodating in the sense that the slop (referring to text written by a machine, at the moment of this writing) sounds imperfectly human, just as this piece of text.We have chosen accommodation.
[02] The Methodology
The goal is to build a machine that doesn't just "generate" text but actually hesitates over it— a system that captures the urge to write content that feels personal, perhaps a trait from being a writer before the world fell in love with efficiency. We leave the "imperfections" totally unedited because that is where the human author lives and where the human "you" would read it.It is a humble wave of joy in being imperfect and, ironically, making the AI as imperfect as possible.
[03] The Protocol
We suggest— or are implementing as of now— a few preliminary steps in achieving the above-mentioned methodology.1. Generative Neurosis: A proprietary LLM fine-tuned not on accuracy, but on the cognitive friction of human thought. It produces text that is "beautifully flawed" by default.2. The Rewriter Engine: A stateless input/output gate where users submit sterile, synthetic text and receive a "humanized" version.3. Intrinsic Latency: There is no human in the loop. The hesitation, the syntax errors, and the awkward phrasing are generated entirely from within the machine — a mathematical simulation of a writer's block.