Zenn essay: bad AI writing is usually the prompter's fault

The essay is framed as a corrective to complaints that surface on X and internal Slack about AI output being poor or slides looking slightly wrong. The author's comparison is to AI coding: early attempts with casual prompts broke specs or introduced vulnerabilities, and teams responded by writing better tickets and materials, which is the same discipline he wants applied to writing.

His key distinction is that people bring unconscious, unwritten domain knowledge — the things never put in a README or spec — while AI works only from documents and articles it can access. That gap, he says, explains the sense of wrongness readers feel in generated text.

He is candid that his own practice is still being worked out, with the most useful preparation not yet settled. The piece ends with a note that it was drafted by a human and then reviewed and lightly adjusted by AI, which he offers as a small demonstration of the human-led, AI-assisted balance he recommends.