Every story we've covered involving anthropic-opus.
A Graphite study comparing AI-written rewrites of pre-ChatGPT articles with human originals finds thousands of recurring phrase-level ‘tells’ across leading large language models. The analysis highlights model-specific habits — for example, Anthropic’s Opus 5.5 overuses words like “dependable” and the phrase “this matters,” while OpenAI’s Astra favors corrective and hedging constructions — even as common early giveaways such as em-dashes decline.
No stories here yet.