Notes from building agentic systems, tuning local LLMs, and working seriously with AI.
-
Proving Coverage With an Agentic SOC
When detection logic becomes probabilistic, proving security coverage may depend less on owning static rules and more on continuously validating what the system can actually detect.
-
My AI SOC Manager Earned Its Keep. The Debate Round Didn’t
A side-project agentic cybersecurity team that reviews real security evidence, produces auditable findings, and hands the important work to humans — augmentation, not replacement.
-
Standing up the blog: repo as source of truth
Why this site is markdown in a git repo, built with Astro, and deployed to Cloudflare Pages — and how posts actually get published.
-
My Agentic SOC Scored 100% on My First Test. That's Exactly Why I Didn't Trust It.
A perfect score on the first planted-threat test hid the most valuable finding: an AI analyst that spots a real threat, then rationalizes it away — and a manager that signs off without questioning it.
-
I Built an Agentic Cybersecurity Team Because I Got Tired of Waiting for One
A side-project agentic cybersecurity team that reviews real security evidence, produces auditable findings, and hands the important work to humans — augmentation, not replacement.