Home  /  Uses

Uses

The things I actually reach for, not the things I have installed. If it is on this page I have used it in the last month.

Machine & shell

OS
macOS on Apple Silicon. Homebrew under /opt/homebrew. Everything that can be installed with brew is.
Shell
zsh — with bash as the assumed lowest common denominator for anything scripted that another human will run. Scripts get #!/usr/bin/env bash and set -euo pipefail, because a script that keeps going after a failed command is a script that lies to you.
Editor
VS Code. Not because it is the fastest, but because the YAML, Terraform and Kubernetes extensions are the ones that get schema updates first, and I spend more time in schemas than in keystrokes.
AI
Claude Code — in the terminal and in the editor. Most useful for the second draft: reading unfamiliar code, drafting the boring 80% of a composition, and arguing with me about a design before I commit to it.

Platform & infrastructure

Control plane
Crossplane with function-pythonic compositions, and Upjet-based providers when a Terraform provider already solved the resource properly. Kubernetes as the API for infrastructure, not just for containers.
IaC
Terraform and Terragrunt — still the right answer for plenty of things, and the thing I most often migrate from. Burrito when Terraform needs to live under GitOps rather than on someone's laptop.
Delivery
ArgoCD and FluxCD, app-of-apps, with Argo Rollouts for canary and blue/green and Argo Image Updater to close the loop on tags.
Kubernetes
EKS, Helm, Karpenter for node lifecycle, Cilium and the Gateway API for networking. Consolidation on, PodDisruptionBudgets written like you mean it.
Secrets
External Secrets Operator, HashiCorp Vault, SOPS for things that have to sit in a repository. AWS Pod Identity over anything long-lived.
CI
GitHub Actions and GitLab CI, with OIDC into AWS. I have maintained enough Jenkins to be genuinely pleased not to be.
Watching it
Datadog and Grafana day to day; Splunk and New Relic where I have inherited them.

Languages

Daily
Bash / zsh and YAML. The honest answer for this job.
Working on
Python and Go. Python is where compositions get interesting; Go is where the operators I depend on are written, and reading them is the fastest way to stop guessing what a controller does.

This site

One HTML file with inline CSS, a shared stylesheet for the secondary pages, no framework, no build step, no third-party requests. Static objects in an S3 bucket behind CloudFront, deployed by a GitHub Actions workflow that assumes a role via OIDC — there are no AWS keys in the repository or in the Actions secrets. Infrastructure is Terraform, in the same repository.

The whole page is about 25 KB before compression. It is a résumé; it does not need to hydrate.