A new paper on arXiv argues that current vulnerability discovery methods are insufficient for today's large and complex software systems. The authors contend that scalable and autonomous techniques are needed, but existing approaches still rely on human domain knowledge and ongoing human intervention throughout the process.

The paper's central claim is that tools alone are not enough. Instead, it calls for bringing agents into the hunt—autonomous systems that can take a more active role in finding vulnerabilities. While the abstract is brief, the direction is clear: security research may need to shift from human-guided tool use toward agent-driven discovery.

This is a single-source summary, so no independent comparisons are available. The significance lies in the proposed move beyond conventional tooling toward greater automation in security analysis.