Asking LLMs questions to help security researchers quickly understand the functionality of code, using LLMs to quickly generate test code, integrating LLMs into IDEs to provide coding security suggestions, etc. In 2024, to push the boundaries of LLMs in automated vulnerability discovery, teams in the AIxCC competition, Google, are trying to use AI agents, combining traditional vulnerability analysis methods, adding more tools to LLMs, and guiding the models to reference human research methods to conduct more autonomous vulnerability analysis.