Project Glasswing: What Mythos Shows Us

玻璃翼项目：Mythos向我们展示了什么

The team applied Mythos and other security-focused LLMs to real-time code testing for critical infrastructure; the post reports vulnerability detection strengths, false positives, and unstable context handling, but does not disclose sample size or benchmark metrics.

Why it matters: Cloudflare offers applied observations on security LLMs testing critical-infrastructure code, so HKR-K/R pass. Missing sample size and metrics keep it low in the 72-77 band; no hard-exclusion rule applies.