Reddit user criticizes Anthropic safety approach

User claims Anthropic's safety model has ~99% false positive rate and routes queries through Opus at double cost. The post argues ChatGPT offers better performance without such heavy guardrails.