case study
ThreatBase: Agentic Intelligence for OSINT
Open-source intelligence (OSINT) is powerful but fragmented. ThreatBase was built to scale it with agentic workflows, turning disparate sources into a structured, queryable system. It's the same thesis I apply to portfolio companies like Skydio and Swan: build infrastructure that compounds; let the system handle throughput.
Problem
Analysts drown in signals. News, social, dark web, corporate filings, the sources multiply. Manual collection and triage don't scale. By the time you've synthesized a threat landscape, it's stale. The gap between data and decision keeps widening. Every new source adds manual work instead of adding to the system.
I've seen the pattern: analysts spend 80% of their time collecting and normalizing, 20% on actual analysis. That ratio is backwards. The job is judgment and insight, not copying data between tools.
Approach
Agentic intelligence infrastructure: agents that ingest, normalize, and structure OSINT. Automated workflows for collection, deduplication, and enrichment. A layer that turns raw signals into monitorable entities and actionable insights. Humans stay in the loop for judgment; the system handles the throughput. Same pattern as Skydio, intelligence in the system, not the operator.
The key: structured output. Raw feeds become entities, relationships, and timelines. Queryable. Alertable. The analyst asks "what changed?" instead of "where do I look?"
Outcome
Analysts monitor for threats at scale instead of manually chasing leads. Faster detection. Better coverage. The infrastructure compounds, every new source and workflow adds to the system instead of multiplying manual work. It's OSINT, evolved for an agentic world. And it's the same lens I use for all my infrastructure bets: build once, deploy everywhere.