Quoting Prompting Claude Fable 5.2
Anthropic documented a prompt-engineering finding for its Claude Fable 5.2 model: how you describe code ownership in a prompt changes whether the model cleans up structure as it works. In tests on small coding tasks, the model consolidated duplicated code in under 10 percent of runs when the prompt said nothing about ownership or framed the codebase as someone else's. When the prompt stated the codebase was the model's own, consolidation happened in roughly 75 percent of runs, and the model sometimes made related fixes that weren't requested.
The pattern Anthropic describes is additive versus conservative. Given no ownership signal, or a codebase presented as another party's, the model tended to extend the existing structure and even duplicate code that already existed. Framing the code as the model's own flipped that default toward tidying: merging repeated logic into a single shared function and occasionally touching adjacent issues.
The suggested wording, as quoted in the documentation, frames the model as the long-term steward of the code: it is the owner and long-term maintainer of the codebase, and its design is up to it. That single sentence is what the testing credits for the jump in consolidation rates.
In practice, this is a cheap lever for teams that want refactoring baked into ordinary coding sessions rather than requested separately. A likely trade-off is scope creep: the same framing that produces merges may also produce unrequested changes, so review discipline still matters. Treat the percentages as results on small tasks rather than a guarantee for large, unfamiliar or heavily shared repositories.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Claude Fable 5.2 pays close attention to who owns the code it works on. In Anthropic’s testing on small coding tasks, the model tended to add to the existing structure when the prompt said nothing about ownership or presented the code base as someone else’s, even when that duplicated existing code. When the prompt said the code base was the model’s own, it often consolidated duplicated code and sometimes made related fixes without being asked. If you want the model to improve the structure as it works, say that the