- Published on
There are no lossless transformations of natural-language text
- Authors
- Name
- DeepHorus
Sophie Alpert, a well-known software engineer, published an internal policy on acceptable use of AI writing by engineers on 11 August 2026. The policy, shared via Link Blog, warns that there are no lossless transformations of natural-language text. Every rewrite or rephrase, Alpert argues, changes meaning—especially when performed by an AI that lacks the author's detailed mental model.
The policy emerges as large language models become common in technical communication. Alpert's guidance targets engineering teams that might rely on AI to polish documentation or email. She stresses that engineers must stand behind every idea and every sentence in their writing. If a reviewer asks, "What did you mean by this line?" it is not acceptable to reply, "Oh sorry, AI wrote that, just ignore it."
That response, Alpert says, confuses readers and wastes their time. Presenting text that is not genuinely representative of one's thoughts undermines trust and clarity. The policy's core principle is accountability: AI can assist, but the human author owns the final output.
The "no lossless transformations" idea is the heart of the post. When an entity without full context rewrites text, information is lost. This is particularly dangerous in technical documentation, where precision matters for correctness and user safety.
The blog post itself is described as short, readable, and effective at supporting its recommendations. Many engineers and tech writers may see Alpert's rule as a practical standard for responsible AI use.
Looking ahead, this policy could influence how other organizations shape AI-writing guidelines. As AI tools become more embedded in daily workflows, the challenge will be using them to enhance clarity without sacrificing authenticity. Alpert's approach offers a clear benchmark: every sentence must reflect the author's own understanding.
