1970 exploitable findings later.

Modern vulnerabilities live in the seams between systems. Here's how reasoning-based AI security catches what static analysis misses, and where it still gets it wrong.