all public benchmark results and user feedback paint a quite different picture. Chinese have coding agents on par with Claude Code, they could easily FT/RL to future improve its specific capability if they want, yet anthropic refuses to even acknowledge the reality.