Sophie Alpert argues that there are no lossless transformations of natural-language text. Her recent post outlines an internal policy for engineers using AI to write documentation. She insists that developers must stand behind every idea and sentence in their final output. If a reviewer questions a specific line, the author cannot dismiss it by claiming an artificial intelligence generated it. This approach prevents confusing readers with content that does not genuinely represent the writer’s thoughts.
The core issue is that every rewrite or rephrase alters the original meaning. When an entity without a detailed mental model of the author’s intent performs the rewriting, information is inevitably lost. This applies specifically to technical documentation where precision matters. The policy aims to ensure accountability and clarity in professional communication.
* AI tools cannot fully capture the nuanced mental model of the original author.
* Reviewers must be able to trace every claim back to the human writer.
* Disclosing AI usage for specific sentences creates confusion rather than transparency.
