Thomas Ptacek argues that frontier AI models are about to fundamentally disrupt vulnerability research and exploit development. LLM agents are uniquely suited to finding zero-days due to their encoded knowledge of bug classes, pattern-matching across large codebases, and ability to run brute-force reachability analysis. The economics and practice of the entire field are expected to shift within months, not years.
Research
Vulnerability Research Is Cooked
Frontier AI models will democratize zero-day discovery through automated code analysis and reachability testing, collapsing the economic moat of vulnerability research within months.
Saturday, April 4, 2026 12:00 PM UTC2 MIN READSOURCE: Simon WillisonBY sys://pipeline
Tags
research
/// RELATED
Infrastructure4d ago
On Dwarkesh Patel's Podcast With Nvidia CEO Jensen Huang
Zvi critically examines Nvidia CEO Jensen Huang's credibility on semiconductor and AI infrastructure claims, distinguishing his narrative discipline from other executives who make provably false statements.
Research1d ago
Cloud Is Closer Than It Appears: Revisiting the Tradeoffs of Distributed Real-Time Inference
New analysis challenges assumptions that edge-only is optimal for real-time ML inference, showing cloud and edge have more nuanced tradeoffs depending on latency, bandwidth, and cost constraints.