Hi, I'm
Mike Roberts
cloud security · ai · builder
I lead a team of Customer Engineers at Google supporting Wiz, the cloud security platform Google acquired earlier this year. I've worked in technology for two decades, with a focus on cloud security for the past several years.
Outside of work, I use AI coding tools to build apps such as a privacy-first interval timer called TrainIQ and a strength training app called StrengthIQ. Both are built around a simple rule: your data stays on your device.
I write about AI, security, and the patterns I observe at the intersection of both. I'm increasingly interested in teaching — specifically at the intersection of AI and cybersecurity, where I think the knowledge gap is widest and growing fastest.
What I Build
Apps for people who take privacy seriously
Writing
Notes on AI, security, and building things.
Building an OpenClaw Agent with Memory, Tools, and Trusted Workflows
A practical walkthrough of how I set up OpenClaw as a personal AI operating system, what choices mattered, and why useful agents need more than a good model.
Securing a Personal AI Agent with Hardening, Secrets, Boundaries, and Approvals
A practical security deep dive on running a personal AI agent with real tools, local memory, secrets, approvals, and workflow boundaries without turning it into a liability.
Supervising AI While Building the Platform You Write On — Lessons from an Astro + TinaCMS Migration
A practical account of using Claude Code to migrate a personal site to Astro and then layer on a TinaCMS publishing workflow, with notes on what the AI got right, what it got wrong, and why the muscle of supervision matters for AI, cybersecurity, and teaching.