This is a free preview of a subscribers-only post.
Was this newsletter forwarded to you? Sign up to get it in your inbox.
Claude Opus 5 is a hard model to love. In its first week at Every, it argued with instructions, stopped before the work was finished, and fought the skills and plugins we built for earlier Claude models.
Then we deleted those instructions.
With less process, Opus 5 sometimes got dramatically better. It built strong software, worked through bugs for hours, and produced rigorous writing and knowledge work. Low and medium thinking effort also brought out more of its strengths in coding while reducing some of its annoying quirks.
The reversal left us with a harder question than whether Opus 5 is good: Is the model disappointing, or are we judging it with systems built for its predecessors? We get into: