// Blog
What I've Learned
Reflections on building, shipping, and figuring it out as I go.
November 2, 2026

Detecting Okta Configuration Drift With GitOps

A screenshot proves what a setting said on one day. This post describes how to read a system’s configuration every day, compare it with settings a person approved in Git, keep the result as audit evidence, and alert on differences. Okta is the example, and build or buy is the lens.

Continuous Monitoring GitOps Build vs Buy GRC
Read post →
November 1, 2026

What I Learned Building Programs With Claude (Part 5 of 5)

Building programs with Claude changed how I work. This post collects what held up across the previous four posts and what remains open.

AI Agents Program Management Lessons Learned Claude
Read post →
October 23, 2026

Design Choices for Agents That Touch Real Work (Part 4 of 5)

An agent that reads records carries one set of risks. An agent that changes them carries another. This post covers the design choices Claude and I use: report-only, propose-only, independent review before execution, and an iteration cap.

AI Agents Agent Design Program Management Claude
Read post →
October 13, 2026

Nothing Goes Stale (Part 3 of 5)

Work goes stale in the gaps between reviews. This post compares continuous and periodic review, and explains why cadence and ownership hold a program together more than the tools do.

Program Management Review Cadence Ownership GRC
Read post →
October 7, 2026

Patching Dependencies with a Seven-Stage Agent Pipeline

Identity-Aware Proxy covers who can reach an internal app. Dependencies enter through a different path. This is how Claude Code and a set of independent agents keep those packages patched, with a person approving every consequential step and a rollback ready at each one.

Vulnerability Management AI Agents Dependabot Claude Code
Read post →
October 6, 2026

Building a Program From a Blank Page (Part 2 of 5)

Every program I have started began as a sentence I could not yet turn into a process. This is how Claude and I move from a vague need to something that runs on a cadence without me starting it.

Program Management AI Agents Process Design Claude
Read post →
September 29, 2026

The Scrum Master for My Agents (Part 1 of 5)

Every agent I built solved one problem and created another: I could no longer tell which ones had stopped working. This is how Claude and I built a report-only agent that runs a daily standup over all of it.

AI Agents Program Management GRC Claude
Read post →
September 15, 2026

I Wrote the Backup Guide. Then I Failed My Own Audit.

In July I published a checklist for backing up a local AI agent setup. Seven weeks later my laptop died. This is me grading my own homework, which turned out to be a less satisfying exercise than writing it.

Disaster Recovery Postmortem AI Agents Backup
Read post →
July 27, 2026

The Backup Plan Nobody Writes for Their AI Agent

We internalized the drill for our data years ago. Then we started building agents that run locally, and quietly recreated a single point of failure we thought we had retired.

AI Agents Disaster Recovery Backup Cowork
Read post →
July 24, 2026

Your Risk Register Is a Spreadsheet. Your Risk Program Doesn't Have to Be.

Security risk management at most organizations lives in a spreadsheet someone built three years ago. The spreadsheet works. The process around it does not. Here is how I automated the mechanical layer while keeping the spreadsheet as the source of truth.

Risk Management GRC Automation Apps Script
Read post →
June 25, 2026

Four Controls Between an AI Agent and Everything It Can Reach

OWASP and Docker published separate guides on securing AI agents within weeks of each other. They come from different angles and land in the same place. The overlap is the playbook.

AI Agents Agent Security OWASP IAM
Read post →
June 21, 2026

I Reverse-Engineered a Weekly Security Report and Accidentally Automated a Program

It started with a screenshot of a manual Friday report. It ended with a system that handles the operational memory behind an entire vulnerability management program.

Vuln Management Automation GRC
Read post →
June 11, 2026

Your AI Agent's Backend Is Everyone's Backend

The Model Context Protocol connected AI to your internal tools. It also connected your internal tools to every vulnerability class from the last twenty years of application security, all at once.

MCP Security AI Agents API Security OWASP
Read post →
June 5, 2026

Someone Built a Coding Agent on Chipotle's Dime

An unauthenticated customer support chatbot became a free AI coding assistant. Here is what went wrong, what the exploit looked like, and what every company deploying AI endpoints should do before the same thing happens to them.

AI Security API Gateway Authentication
Read post →
May 23, 2026

TIL: What Is Refactoring (and Why Your Vibe-Coded App Needs It)

I built an internal tool with AI. It works, it is in production, and people use it every day. Now I need to reorganize the code so it can keep growing. Here is what refactoring means and how to do it without breaking everything.

Refactoring Vibe Coding AI Development
Read post →
May 9, 2026

Ship Your Portfolio This Weekend

A domain, GitHub Pages, and Claude. Here is the playbook for building a portfolio site from scratch, what it costs, and why practitioners in GRC and security should have one.

Portfolio GitHub Pages Claude
Read post →
May 3, 2026

What Was Once a Dream Is Now a Reality

I went from never having shipped an application to running one in production in six days. This is the story of building a TPRM platform with Claude, and why the bottleneck between knowing what to build and being able to build it is gone.

AI Development TPRM Claude GCP
Read post →