The AI slop problem nobody is talking about

Nate B Jones speaks with Chris Best, co-founder and CEO of Substack, about what AI slop really means for writers, readers, and online communities. Best frames slop broadly: not simply AI-generated text, but content that nobody truly stands behind — spam, clickbait, automated replies, or plausible prose produced without intent.

For Substack, the mission is almost the opposite. The platform is meant to help people make things they believe in and earn a living from that work. That does not make the stance anti-AI. Best and Jones both treat AI as a powerful creative tool when it is used to clarify ideas, edit prose, or accelerate a thoughtful process. The problem begins when the tool is used to counterfeit effort, automate engagement, or hide the absence of a real authorial commitment.

A central distinction in the conversation is that the question should not only be where the words physically came from. The deeper question is whether a person actually thought through, chose, edited, and took responsibility for the piece. Jones describes his own use of LLMs, Whisper Flow, Codex, and Claude as an adversarial editing loop: getting ideas out, pushing the model, removing familiar LLM habits, and making the final writing more precise and human.

Pangram comes up as one possible signal for identifying certain kinds of misuse, but not as a perfect or final answer. Best emphasizes that Substack wants an open conversation with creators because the real answer will be cultural as much as technical. Norms around AI-written comments, AI-assisted drafts, video avatars, and conversational agents will have to become much more explicit.

The strongest claim is that being anti-slop can be a pro-AI position. If AI is going to be integrated into healthy human communities, the systems around it need to protect trust, effort, voice, judgment, and disclosure.

Key takeaways

• Slop is best understood as content without genuine intent, not merely content made with AI.
• Substack’s position is anti-deception and anti-spam, not anti-AI.
• The important boundary is responsibility: did a human think, edit, and stand behind the work?
• Detection tools may help, but norms and expectations will matter just as much.
• As content production gets cheaper, trust and human attention become more scarce and valuable.

Source

• Chaîne: AI News & Strategy Daily | Nate B Jones
• Vidéo source: https://www.youtube.com/watch?v=m_ZyTNmCDeY