Depthfirst AI finds critical web vulnerabilities Mythos missed at 1/10 cost

Depthfirst, an AI cybersecurity startup, claims its specialized model has uncovered critical vulnerabilities that Anthropic's broader Mythos system missed—at just one-tenth the cost. According to CEO Qasim Mithani, Depthfirst's AI optimizes for the single task of vulnerability discovery, allowing it to achieve a complete audit for $1,000 compared to Mythos's $10,000. The discovered bugs are significant enough to impact the majority of people using the web today.

Mithani argues that general-purpose AI models like Mythos, while powerful, lack the efficiency of purpose-built systems. Depthfirst's focused approach sacrifices breadth for depth, enabling it to identify issues in common web infrastructure that broader tools overlook. The cost difference is dramatic: what Mythos does for $10,000, Depthfirst can do for $1,000. For enterprises, this makes continuous, affordable security auditing feasible. The findings suggest a shift toward specialized AI agents in cybersecurity, where narrow expertise can outcompete general intelligence on specific tasks.