AI Slop

A teammate asks an AI tool to “write a blog post about our product,” skims nothing and publishes it. The result is 800 words that sound plausible, repeat the same three points and state one feature that does not exist. Nobody asked for it, nobody checked it and every reader pays a little time to discover it was empty. That is slop. Compare a writer who uses AI to draft, then checks the facts and cuts it down: same tool, different result.

Simon Willison, who helped popularize the term in 2024, draws that line explicitly: not all AI-generated content is slop, and the label fits content that is mindlessly generated and pushed at people who did not ask for it (Willison). Merriam-Webster made “slop” its 2025 Word of the Year, defining it as low-quality digital content “produced usually in quantity” by AI (Merriam-Webster). The phrase “in quantity” is the important part. Generative AI makes producing content almost free, so the cost moves to whoever has to read, review or filter it.

On the show, the cost shows up most clearly in software. In episode 74 at 27:06, Laurie Voss says open-source maintainers felt it first: “we are drowning in slop,” with more AI agent pull requests than anyone can review. Episode 74 names Inngest, Svix, Walrus Memory and G2i as the season’s presenting sponsors; the read is at 4:38. In episode 56 at 45:52, AMD’s Anush Elangovan says the early reaction to AI-written commits was “AI slop, AI slop,” and the fix was to require tests to pass before a person looks at the code. Episode 56 is sponsored by Galileo, the Season 3 presenting sponsor; the read is at 32:55. In hiring, Greenhouse CEO Daniel Chait describes in episode 58 at 15:35 how the “noise and chaos and AI slop” of mass applications gives fraudsters cover, including the deepfake candidates covered on that page. Episode 58 is sponsored by Galileo, the Season 3 presenting sponsor; the read is at 2:25.

Our synthesis from those examples: the remedy is rarely banning AI. It is putting a check between generation and the audience, such as automated tests, a human-in-the-loop review, or AI evaluation that rejects weak output before anyone else has to read it.