Science fiction author Charlie Stross published an essay explaining why he uses no LLMs anywhere in his fiction writing or personal life, criticizing vendors who market chatbots as substitutes for verifiable research rather than generators of “plausible answer-shaped text.” He argues that fiction resists the kind of AI assistance that works reasonably well in software development, since prose relies on long-range narrative dependencies that constrained programming languages and APIs do not. Stross notes he sometimes leans on deliberate novelty — such as setting a fantasy novel among Bronze Age Alaskan god-kings — partly to make any AI involvement in his process implausible, and points to Advait Sarkar’s essay “AI Should Challenge, Not Obey” as a framing he finds more useful than most industry commentary.
