How we run swarms of AI hacking agents against ourselves in a game of capture the flag

How Lovable built an internal offensive security program around verifiable results, and how we orchestrate a swarm of AI hacking agents to find real vulnerabilities instead of noise.