The AI Slop Problem: Intent vs. Automation

The Rise of AI Slop and the Erosion of Trust

Substack CEO Chris Best and host Nate B. Jones define "slop" not merely as AI-generated text, but as content created without human belief or intent. The core problem is that the internet is becoming inundated with low-effort, automated content that functions as a denial-of-service attack on the public square. When readers can no longer distinguish between human-authored thought and machine-generated output, the value of all online discourse diminishes, leading to a breakdown in discovery and community engagement.

The Limits of Detection

Substack has introduced an integration with Pangram, an AI detection tool, to provide users with transparency regarding whether a text was generated by an LLM. However, both Best and Jones emphasize that detection is not a moral judgment. A piece of writing can be "AI-generated" according to a detector but still represent deep human thought and iteration. Conversely, a human-written piece could be low-quality, and an AI-generated piece could be high-quality. The tool is intended to spark conversation and provide a signal, not to police content or restrict the use of AI tools.

The "Proof of Work" Crisis

Historically, the act of writing a long-form piece served as a "proof of work"—a signal that a human had invested time and attention into a topic. As AI makes the production of plausible-sounding text nearly costless, this signal is effectively dead. The challenge for creators is to move beyond the "average distribution" of ideas that LLMs naturally gravitate toward. Because models are trained on consensus data, they often default to generic, safe, or repetitive arguments. True human value now lies in the ability to push against these defaults, iterate through multiple drafts, and maintain a unique, high-context perspective that the model cannot replicate on its own.

Redefining Human-AI Collaboration

Jones describes his own workflow as a "wrestling match" with AI, where he uses tools like Claude and Codeex not to generate final products, but to iterate on ideas. He treats the AI as a sounding board, pushing back when the model's output becomes too cautious or generic. This process requires the human to maintain a clear vision of the thesis, using the AI to refine the prose while ensuring the final output remains bold and intentional. The goal is to move from "prompting for output" to "co-thinking," where the human remains the primary driver of the conceptual direction.