Recent advancements in AI have led to the development of Aletheia, a math research agent designed to automate the process of mathematical research. Aletheia can generate, verify, and revise solutions using natural language. The system is built on an enhanced version of Gemini Deep Think, which allows it to tackle complex problems beyond typical competition-level challenges. This tool not only processes Olympiad-level problems but also engages with higher-level academic tasks. The article highlights several significant achievements of Aletheia. One notable accomplishment is the generation of a research paper, Feng26, which calculated structure constants in arithmetic geometry without human input. Another paper, LeeSeo26, showcases a collaborative effort between humans and Aletheia, proving bounds on independent sets in particle systems. Furthermore, Aletheia evaluated 700 open problems related to Bloom's Erdős Conjectures, autonomously solving four of these questions, marking a substantial step in AI's role in mathematics. To enhance public understanding of AI's capabilities in mathematics, the authors propose measuring levels of autonomy and novelty in AI-generated results. They suggest introducing "human-AI interaction cards" to improve transparency regarding the involvement of AI in research outputs. The paper emphasizes the importance of human-AI collaboration and provides access to all prompts and model outputs, allowing for deeper exploration of these findings.