Why I’m moving to Linux (for real)
TL;DR
APFS is catastrophically slow for developer workloads: Git clean and PNPM install take 35 seconds on a maxed-out MacBook Pro versus 2-7 seconds on a Linux box with ext4, despite Apple's SSD being 2-3x faster hardware-wise.
Mac's security process chokes on AI agents: CIS policyd, Apple's process monitor, spins up to track every subprocess that Codex creates, meaning sub-agents with 30+ processes each can saturate even an M5 Max.
Subscription token subsidies break cloud IDE economics: Claude's $200 plan delivers $8,000/month of inference value, but you can't use those subsidized tokens in Devon or Cursor cloud, making local agents with remote machines the better play.
Network KVMs give AI full hardware control: The GL.iNet Comet Pro plugs into HDMI and USB, letting Theo's agents control bootloaders, flash OSes, and fix broken Linux installs remotely, with a Fingerbot for physically pressing power buttons.
Linux boxes idle at 0.1-0.2 vCPUs: While his Mac fans screamed, his 32-thread Linux machines barely break a sweat running multiple sub-agents, letting him queue multi-hour agent tasks without tethering himself to an open laptop.
Summary
Theo abandoned his maxed-out MacBook Pro for Linux machines after watching his CPU hit 100°C from running AI coding agents, discovering that APFS takes 35 seconds for git operations that Linux handles in 2 seconds. He now runs a fleet of Linux boxes controlled remotely through Tailscale, network KVMs, and his own T3 Code app, letting agents work for hours while his laptop stays closed.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Retirement Email Isn't a Warning
Model retirements now arrive every few weeks; the config-eval-rehearsal loop turns each deprecation email from a fire drill into an afternoon swap.
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.