What happened
AWS detailed (Oct 5) that Continuum, its multi-agent vulnerability service, passed 819/920 CyberGym-E2E tasks (89.0% end-to-end on the main S3 measure) — 23.1 points above the prior public high of 65.9%, finding, demonstrating and tested-repairing vulnerabilities across 139 OSS-Fuzz-based projects, rising to 93.7% beyond the 90-minute limit.
Why it matters
It is a strong public benchmark disclosure that reframes what autonomous code security can achieve end-to-end (discover→prove→repair), a capability benchmark other autonomous-patching vendors will be measured against.
Applicability
Teams running or piloting autonomous vulnerability management should weigh Continuum's measured E2E capability against internal triage loads; use the public benchmark for vendor comparison.