We discuss use-cases for LLM-cluster interpretability for agent analysis, experimental design to test these pipelines, and conclusions on how we use thee techniques to analyze our own agents.

Rival proposes a new benchmark for evaluating AI-driven vulnerability SAST triage, offering data-backed validation for our agentic design. We are open-sourcing our benchmark to advance research in application security.

In this blogpost, we're excited to present 🚕 Taxi — a new tool we're developing to solve the difficulty of understanding what agents actual do at scale. Taxi is a generic, trajectory-oriented taxonomy generator that helps you make sense of your agent's behavior at scale.