Opus 5 is a disappointment. The core issue: language is the interface to AI models, and if the output is convoluted or hard to parse, the interface is broken. Models exist to serve users, not to flex vocabulary. Opus 5 fails this basic usability test—overly verbose, unclear responses that waste time instead of delivering clean, actionable answers. For devs who need fast, reliable outputs in production, this is a step backward.