Early Claude Opus 5 tests say it is strong, but breaks older agent workflows
every · x · 2026-07-25
Early testing says Claude Opus 5 is powerful but brittle with existing workflows
The post quotes an internal first look at Claude Opus 5 and describes a mixed reaction after testing it across coding, writing, knowledge work, and an internal agent.
Main takeaways:
- The model argued with instructions, stopped early, and did not fit well with existing skills/plugins such as Compound Engineering.
- After deleting the old skill stack and starting fresh, the model improved dramatically.
- In a simpler setup, it showed flashes of brilliance.
- The main warning is backward compatibility: existing workflows may need to be rebuilt around it.
The quoted launch line says Opus 5 is a thoughtful, proactive model that comes close to frontier intelligence while costing half as much.
Related event: Claude Opus 5 Early Tests: Better Efficiency but Overly Proactive(29 posts)→
More from Models
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Local beats cloud: Qwen 3.8 FN handles private student data as well as frontier models — dr_lm · 2026-10-27
- Aidan Gomez: model's CoT said no tool calls right before making them — aidangomez · 2026-10-27
- Nat Lambert: Cohere's model is narrow RL training on internal infra ill-suited for open weights — natolambert · 2026-10-27
- OpenAI teases DOTS is coming to Europe 'soon' — mark_k · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27