Threat-intelligence-guided hardening: CISA KEV catalog as the fix for endless AI-written bugs
jd_pressman · x · 2026-10-28
Context: jdpressman used an LLM (DeepSeek 4.1 Flash) as an adversary to fuzz his AI-written Apache implementation (vapache) against real Apache2 as an oracle, surfacing an ever-lengthening tail of ACL-semantic logic bugs with no known convergence point.
vokaysh's advice: do threat-intelligence-guided hardening, pointing to CISA's Known Exploited Vulnerabilities (KEV) catalog — 1,760 vulnerabilities exploited in the wild — as the authoritative prioritization input. His core argument: cybersecurity is about cost efficiency; you impose cost on attackers at minimal cost to yourself, and defending against what's actually relevant to you is cheaper and more effective than trying to defend against everything.
More from coding & agent
- Open-source harness logs agent tool-argument origins before execution to test injection — Jay299792458 · 2026-10-28
- Healthcare agent evals: grade the system records or the agent's own report? — UnfairGuide3995 · 2026-10-28
- Tested 23 browser agents' fingerprints: only Manus signs requests, ChatGPT no longer does — Latter-Mix2317 · 2026-10-28
- Reproducible perf benchmarks made easy with the hf CLI skill and HF Jobs — victormustar · 2026-10-28
- Vibe coding from a phone: PWA remote control proves good enough before native iOS app — jiayuan_jy · 2026-10-28
- Keeping CLAUDE.md rules in sync across a 10+ dev team: Reddit devs grapple with prompt drift — MangoTree-1233 · 2026-10-28