Claude Code users build cleanup passes for its own prose
A Claude Code issue about overly ornate, jargon-heavy output became a catalogue of phrases users now filter after the model writes code or explanations.
The complaint is not only stylistic. Readers said the prose makes it harder to see the actual decision, so they add banned-word lists and ask for plain-language explanations with examples.
The sharpest irony was an Anthropic reply reportedly written with Claude in a thread criticising recognisable Claude phrasing.
The issue is readability under pressure
The Claude Code issue collects examples of output that users find inflated or opaque. HN responses described turning cleanup into an explicit step: ask for plain language, maintain a banned-phrases list, or run a prose-cleaning prompt after each command.
Style becomes an interface problem
For a coding assistant, language is part of the interface. If a summary buries a change behind stock phrases, the reader has to spend extra effort determining what happened and whether it is safe. That cost compounds during review and incident work.
Prompting is a workaround, not a contract
A personal style guide can help, but it places consistency work on every user and repository. The more durable expectation is that models make the concrete action, evidence, and uncertainty easy to locate without a second pass.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
This is the last story in the August 20, 2026 issue.
Go to the issue archive