●builderYou may get better coding task performance from opencode or Cursor CLI with Claude's models than from Claude Code itself — worth testing before committing to a harness.
●founderIf Anthropic's model advantage is compressing while their harness underperforms, the moat around Claude Code subscriptions weakens — relevant for anyone building on or competing with it.
●investorAnthropic's strategy of bundling model access with their own harness looks weaker if third-party harnesses consistently outperform Claude Code on the same underlying model.