SemiAnalysis quotes user saying Anthropic models build better long-term codebases

SemiAnalysis, in a post on X, quotes a user who says OpenAI models are token-efficient in the short term but skip follow-up work to finish faster.

The user says Anthropic models produce better and more readable codebases over the long term, and that model choice matters more for token efficiency than the user does. The post reports the user's view; it does not include independent testing or measurements of codebase quality or token use.