AI Under Oath

Software engineer applying years of deployment safety and reliability engineering to AI agents. What fails at scale, why, and the patterns that prevent it. No policy abstractions, just what works from someone who gets paged when things break.