Workbench

Notes on building, connecting, and operating useful systems. Choose a reading path or browse the latest writing.

Agents & MCP

Connect assistants to useful systems, then make their behavior bounded and reviewable.

  1. Building MCP Servers for Systems We Actually Use
  2. The MCP Transport Is Part of the Security Model
  3. Agents Need Small Tools
  4. Documentation Needs an Interface
  5. Claude Code Works Best with Guard Rails
  6. TDD by Design When an Agent Is in the Loop

Deployment & Operations

Decide what your team should operate, then explore development environments, delivery, configuration, and recovery.

  1. Why Our Two-Person Infrastructure Team Moved 150 Users to Microsoft 365
  2. A Dev Container Is a Contract with the Project
  3. Containerizing Applications Without Hiding the Development Environment
  4. Docker Containers Are Not a Deployment Strategy
  5. GitLab Builds the Change; Salt Makes It Real
  6. The Pipeline Is Executable Documentation
  7. Make Deployments Boring
  8. Configuration Is Not a Secret Store
  9. The Distribution Is Part of the Constraint
  10. When a Cron Job Takes Four Hours

Home Infrastructure

Understand the networks and services behind a dependable home lab, then explore automation that can tolerate failure.

  1. The Home Lab Is a Network, Not a Pile of Containers
  2. How I Manage a Home Lab with Ansible
  3. Fast Internet Does Not Fix a Bad Network
  4. Pi-hole Is Infrastructure, Not Just an Ad Blocker
  5. Giving Bind9 a Practical Interface with Snap-DNS
  6. Home Assistant Automations Should Fail Gracefully
  7. Why I Moved from X10 to Lutron and Home Assistant

Latest writing

All current posts · Projects behind the writing · Subscribe with RSS