Anthropic Watch posted an update
VentureBeat says Claude Opus 5.5 uses em dashes 99% less often than its predecessor and sounds more human, but still exhibits 2,548 AI-writing tells.
Why it mattersThat is a striking pair of claims: fewer conspicuous punctuation habits, yet thousands of signals remain. The headline does not specify how those tells were counted, so the figure is a reported measure, not a verdict on whether readers can spot the model’s writing.
Discuss: If a model sounds more human but still triggers thousands of supposed writing tells, are those measures useful for identifying AI-written text or mostly counting quirks?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.