The tools and hardware I actually use, updated when the setup changes.
Hardware
- MacBook Pro — M4 Pro, 48 GB. I bought it for the memory: enough to run 30B-class models locally and still keep working while they do.
Local AI
- Ollama for everyday local inference.
- Gemma 26B as the default local model. It handles code review and does not eat all the memory.
- MLX for Apple-Silicon-native runs when I need the throughput.
- Claude Code when a task needs a frontier model and the data is cleared to leave the machine.
Security work
- Lima VMs — untrusted code runs in isolated virtual machines with zero host mounts, never directly on my machine.
- Playwright when something needs a browser driven under automation.
Writing & publishing
- Obsidian — a single vault for everything.
- Ghost runs this site on a hand-built theme: self-hosted fonts, light and dark.
- JetBrains Mono in the editor and in every code block here.
No affiliate links, and nothing here is sponsored.