Interested in being a guest? Email us at admin@evankirstel.com
AI is getting frighteningly good at finding software vulnerabilities and stitching them into workable exploits, and security teams were already underwater long before the latest wave of zero-days. That reality raises an uncomfortable question: if offense moves at machine speed, what does credible defense look like?
We sit down with Dr. Neha Rungta, Director of Applied Science, Amazon Web Services (AWS) unpack AWS Continuum for Code Vulnerabilities, now generally available, and the practical mechanics behind “autonomous security.” We walk through the full loop from discovery to validation to prioritization to remediation, including why proving exploitability in your real cloud environment matters so much for reducing false positives. When you can combine code signals with AWS context like network reachability, permissions, encryption, and workload placement, the output stops being a giant list and becomes an actionable plan.
We also dig into the CyberGym benchmark and what an 89% success rate actually represents: generating a crashing input, producing a patch, preventing regressions by running tests, and confirming the fix resolves the historical vulnerability, not just one symptom. And we get candid about what keeps AI-generated security fixes safe: “boring” but essential engineering discipline, sandboxing, and rigorous test suites.
If you care about secure software development, DevSecOps, vulnerability management, and AI security tooling that engineers will actually use, this conversation delivers real details. Subscribe, share the episode with your security or platform team, and leave a review with the hardest part of vulnerability triage in your org.
More at https://linktr.ee/EvanKirstel
