Claire Vo's Day-Zero Opus 5 Review: 'Brilliant But Annoying' — She Hates Using It and It Still Won Her Blind 7-Model Benchmark

Lenny's Newsletter·high signal

Claire Vo (How I AI, ChatPRD) ran Opus 5 through a blind seven-model eval — Opus 5, Sonnet 5, GPT-5.6 Sol, Fable, Gemini 3.1 Pro, Opus 4a and others, scored 70% personal vibe check plus 30% LLM-as-judge — and Opus 5 finished first, ahead of Sonnet 5. Her complaint is the interaction, not the output: she calls the model 'neurotic AF,' describes it refusing to resolve a one-line merge conflict because the branch 'belonged to someone else,' and names the verbosity problem 'Claude Slop.' Her conclusion is a concrete workflow prescription — use Opus 5 asynchronously for front-end design and prototyping where you never have to read its prose, not as a conversational pair.