1.
projectblack.io | research | | original ↗ | #ai-security | #agentic-ai | #vulnerability-research | #benchmark
The author benchmarks four approaches, including Semgrep and fully agentic GLM 5.1 + Strix, against a known PHPIPAM authenticated LFI (CVE-2026-12194), finding the harness matters more than the model.
Skip to content