Colin Breck rejects AI-generated writing

Generating prose with AI mistakes sheer volume for clarity, creating an absurd loop where developers use LLMs to write bloated documents that recipients must use LLMs to decipher.

• –The context mismatch: Prompt authors already possess the background context needed to navigate AI output, but external readers receive dense, context-free text that demands exhausting line-by-line parsing.
• –Verifying beats generating: Using LLMs to check facts against code, format citations, and catch typos accelerates drafting, whereas asking models to generate explanations from scratch consistently yields superficial, error-prone results.
• –The reader revolt: Developer surveys indicate that nearly 80% of readers abandon AI-scented writing, prompting engineering organizations to establish strict human-authorship requirements for public and internal docs.
• –Compression destroys nuance: Real-world engineering decisions require managing uncertainty and subjective trade-offs that statistical next-token prediction inevitably strips away.