All notes

INSIGHT

Aug 2, 2026

Charles Stross Explains Why He Does Not Use AI in His Writing Process

Science fiction author Charles Stross lays out his reasoning for keeping AI tools out of his writing workflow, a position worth examining for any engineer building AI-assisted creative tooling.

Charles Stross, author of the Laundry Files series and a technically literate voice in science fiction, published a post explaining his deliberate non-use of AI in his writing process. The argument is not a blanket technophobe rejection. Stross engages with software and infrastructure regularly in his public writing. That context matters when reading his position.

The core concern, as understood from his documented stance, centers on what LLM-generated text actually is: a statistical compression of existing human writing, optimized for plausibility rather than originality. For a working novelist whose output is differentiated by voice and specific ideation, that distinction is not abstract. The model produces fluent prose; it does not produce Stross's prose.

There is also a practical integrity argument. Readers of a named author have a contract expectation. When that author publishes under their name, the assumption is that the cognitive work — the choices, the framing, the specific sentence-level decisions — originated with that person. AI assistance, even partial, breaks that contract in ways that are difficult to disclose cleanly.

For engineers building AI writing tools, this is a signal worth tracking. The adoption ceiling in professional creative work is not purely technical. It is reputational and contractual. Tools that treat voice preservation and attribution as first-class features will have different traction than tools that optimize for throughput.

The secondary implication is for training data provenance. Authors like Stross are among the more vocal about where their text ends up. As licensing frameworks for training corpora become more contested, the pool of high-quality, consented professional writing available for future model training narrows. That has downstream effects on model quality in specialized registers.

The post is worth reading directly as a case study in where AI augmentation breaks down for knowledge workers whose output is inseparable from their identity.

Charles Stross Explains Why He Does Not Use AI in His Writing Process | SKYSYNC TECH