Vibe Check: Claude Opus 5 Is Brilliant in Flashes, Frustrating in Practice

Dan Shipper's post points to Every's full review of Anthropic's Claude Opus 5, and the headline sums up its verdict: the model can produce impressive work, but using it day to day was frustrating. The review is mostly behind a subscriber paywall, so this explanation relies only on its publicly visible opening.

That opening gives the reason for the split verdict. During the model's first week at Every, the reviewers say it argued with instructions, stopped before work was finished, and did not fit well with the skills and plugins the team already used, including its compound engineering plugin. The review's framing is that getting Opus 5's best work may require dismantling the systems you already use.

So the complaint is mainly about fit, not raw capability. Anthropic's launch post presented Opus 5 as a model for everyday use that leads several coding and knowledge-work evaluations, while Every's first week with it was marked by clashes with the team's existing skills and plugins. The specific examples and any comparisons in the full Every review are not visible without a subscription.