LLMs didn't kill security testing, they exposed tools that guess at context they don't have. This talk explains what we learned by reinventing program analysis for new end users: builders and agents. We had to develop benchmarks and add determinism with program slicing to better control costs and deliver the optimal price per true vulnerability detected. This presentation is brought to you by Semgrep, an AWS Partner.
Level: 200 - Intermediate
Features: AWS Partners, Lecture-style
Topic: Artificial Intelligence, Developer Tools, Security & Identity
Area of Interest: Application Security, DevSecOps, Agentic AI
Role: Cloud Security Specialist, Developer / Engineer, IT Executive