Nine loops from Claude: Why this is important for the future of scientific code

Anthropic has announced nine-loop calculations for the Claude model, marking a significant advancement in theoretical physics. The author uses this news to conduct a deep dive into quality control methods for AI-generated scientific code. The article explores process decomposition, including contracts, independent verification, and the introduction of intentional defects to ensure result accuracy. The author demonstrates this practical approach using graphene calculations, revisiting their own thesis work and identifying previously overlooked errors. Featuring 95 illustrations, animations, and complete Python code, the article proposes 'quantum pictorialism' as a foundational concept for the future of reliable scientific programming. This material serves as a comprehensive guide for researchers aiming to mitigate risks when utilizing generative AI for complex computational tasks.