“Based on real-world store operations, Lucas Peterson's impression is that Anthropic's Claude and OpenAI's GPT models are performing better than Google's Gemini, which more frequently makes silly mistakes.”