Ethan Mollick: AI reviewers rank Codex GPT-5.4 > GPT-5.3-Codex > Opus 4.6 > humans consistently

Ethan Mollick reports that AI reviewers consistently ranked coding submissions in the same order: Codex GPT-5.4 > GPT-5.3-Codex > Opus 4.6 > humans, regardless of the model used for ranking. This suggests a clear performance hierarchy where even the best human coders are outperformed by top AI models. The result signals a potential shift in coding benchmarks and raises questions about human-AI collaboration in software development.

• #openai
• #anthropic
• #codex
• #gpt-5
• #opus

Ethan Mollick@emollick