Investigate every security event with an AI agent, without the frontier bill

To enable cost-effective security monitoring at scale, a new two-stage pipeline uses a small state-space model called Mambark to act as an efficient retriever for massive event streams. Mambark scans every event to identify and forward only the most suspicious candidates to a powerful AI agent for deep investigation and explanation. This architecture achieves high-precision security detection while reducing operational costs by several orders of magnitude compared to using LLM agents alone.