According to @OpenAI, GPT-5.4 Pro helped solve a 60-year Erdős problem, signaling faster theorem discovery and new math research workflows.

According to Jeff Dean on Twitter, mathematician Michel van Garrel highlighted how Google's latest Gemini Deep Think AI model successfully proved a mathematical conjecture using a novel methodology distinct from traditional human approaches (source: Jeff Dean, Twitter, August 1, 2025). This achievement demonstrates the expanding capabilities of advanced language models in contributing to mathematical discovery and problem-solving. For AI industry stakeholders, this marks a significant leap in leveraging AI for high-value research tasks, opening new business opportunities in AI-augmented scientific research, automated theorem proving, and advanced knowledge generation.