Anthropic’s docs say unstated code ownership pushes the model to extend existing structure and repeat code.

Anthropic’s documentation for prompting Claude Fable 5.2 describes how the model reacts to statements about who owns the code it edits. When a prompt said nothing about ownership, or framed the codebase as someone else’s, the model tended to extend existing structure, even when that meant repeating code.

Simon Willison collected the quotation from Anthropic’s documentation and posted it on 10th November 2026. The testing covered small coding tasks, where the model decides whether to add to an existing code structure or to refactor the repeated parts it encounters as it works.

Builders who want tighter codebases can paste an ownership line into the prompt and tell the model the codebase is its own. Anthropic’s testing found the model merged repeated code in about half the attempts with that sentence, compared with under a tenth of attempts without it.

Watch for guidance covering larger repositories and multi-file changes, since Anthropic tested only small coding tasks. Teams running agentic coding tools should also test whether an ownership statement changes behavior on their own codebases before relying on it.

What matters

• Anthropic reports that prompts with no ownership detail led Claude Fable 5.2 to keep repeating code.
• Builders can lift refactoring rates by telling the model plainly that it owns the codebase it edits.
• Watch whether Anthropic publishes ownership guidance for larger repos beyond small coding tasks.

Why it matters

Watch whether Anthropic publishes ownership guidance for larger repos beyond small coding tasks.

This GenAI News article was prepared in original wording using reporting and materials published by Simon Willison’s Weblog. Source reference: https://simonwillison.net/2026/Nov/10/prompting-claude-fable-52/.

Drafted by the GenAI News review pipeline.