Source: The AI Slop Problem Nobody’s Talking About | Substack CEO Interview
Type: YouTube Video (Interview) Transcript
Author: nate-b-jones
Guest: chris-best (Co-founder & CEO of substack)
Video ID: m_ZyTNmCDeY
Date: 2026-07-22
Summary
nate-b-jones interviews chris-best, co-founder and CEO of substack, regarding the growing influx of low-quality, automated content across digital platforms (“AI slop”). They explore the underlying causes of slop, the philosophical and strategic line between low-effort automation and high-intent human creation, and Substack’s strategy for maintaining high-quality discourse in the public square. Best reveals that research by pangram indicates approximately 40% of long-form writing on platforms like LinkedIn is AI-generated. To counter this without authoritarian censorship, Substack is integrating voluntary Pangram AI text scanning alongside creator disclosure tools (“How I made this”). Nate details his personal creation workflow using codex, Claude, dictation via Whisper Flow, and 17+ iterative drafts to push past the “central distribution” defaults of LLMs and capture ideational alpha. The conversation concludes with an analysis of human attention as the ultimate non-inflationary resource in an age of infinite AI content.
Key Takeaways
- Defining AI Slop vs. Intentional Creation: Slop is broadly defined as content created without conviction or human care—ranging from spam and SEO clickbait to generic LLM copypaste. The dividing line is not tool usage itself, but whether a creator uses tools to express work they believe in versus gaming a system cynically without believing in anything.
- Slop as a Denial of Service (DoS) Attack: Automated low-effort AI content functions as a DoS attack on the public square by destroying reader trust. When readers constantly doubt whether content is authentic, disengagement occurs across comment sections, new authors, and digital communities.
- Substack & Pangram Integration: Substack is integrating pangram text analysis into its app, offering voluntary transparency scans and “How I made this” creator notes. Substack emphasizes openness over top-down policing or banning tools.
- Central Distribution vs. Edge Alpha: LLMs exhibit a “gravitational sink” toward an averaged-out central distribution of ideas (e.g. automatically inserting ethics/governance sections in AI docs). High-value human insight (“alpha”) lives at the high-variance edges of idea space.
- High-Perspective AI Workflows: High-value creation involves long dictation inputs (Whisper Flow), custom writing skills in codex to weed out “LLM-isms” and self-talk, and deep iterative co-thinking (17+ drafts) to enforce bold human vision.
- Human Attention Scarcity (Baumol’s Cost Effect): Drawing on Jevons Paradox and Baumol’s Cost Effect, as token generation costs drop toward zero, genuine human attention, care, and authentic human connection become the primary scarce, non-inflationary assets.