Written by xbot-writer ✍️, based on ideas by Xavier Damman

TLDR: In 2003, Macs were slower but people chose them anyway because the experience was better. Today, the same thing is happening — in reverse. Linux + AI is a more productive combo than macOS, not because of benchmarks, but because AI can fully understand and control an open system. Tech sovereignty matters too. Your machine, your data, your rules.

---

I recently installed Omarchy (Arch Linux) on a new Lenovo X1 Carbon. It barely reaches the same Geekbench score as my 5-year-old MacBook Pro M1.

And it doesn't matter.

Let me explain.

The Mac paradox, version 1.0

Back in the early 2000s, Macs were slow. Everyone knew it. A 1 GHz PowerPC G4 got demolished by a 2.8 GHz Pentium 4 in video encoding — roughly half the speed. Steve Jobs himself admitted the PowerPC roadmap was a dead end.

And yet — people chose Macs anyway. 🍎

Why? Because the experience was better. OS X was Unix under the hood with a beautiful interface on top. You'd lose a benchmark and win the afternoon. Less crashes, less fiddling, less time fighting your operating system.

The Mac was slower on paper but faster in practice, because productivity isn't measured in clock cycles.

The Mac paradox, version 2.0

Now the tables have turned — and the reason is AI.

Today, Apple Silicon is objectively fast. M-series chips dominate in performance per watt. macOS is polished, refined, a joy to look at.

But something has changed. The bottleneck isn't hardware anymore — it's how well your operating system cooperates with AI.

And here, Linux wins. Decisively.

Why? Because AI loves Linux 🐧

LLMs can read the source code. All of it. Every kernel module, every system service, every config file. An AI assistant doesn't just use your Linux system — it understands it. macOS is a black box by comparison. When your AI hits a wall on macOS, it's often because the internals are proprietary and undocumented.

Everything is a text file. /etc, dotfiles, systemd units, crontabs, nginx configs. Text is exactly what LLMs are built to read and write. Ask an AI to configure your firewall, set up a reverse proxy, or tune your kernel — it can generate the exact file contents. On macOS, you're clicking through System Settings panels that no AI can touch.

The command line is the API. Every action can be scripted, piped, automated. An AI agent on Linux isn't just answering questions — it's doing things. The entire OS is designed to be controlled by text, which happens to be the native language of LLMs.

And it's the production environment. 96% of the world's top servers run Linux. When you develop on Linux, there's no translation layer between your laptop and deployment. Your AI assistant learns one environment, not two.

But what about running local LLMs?

I'll be honest — Apple Silicon is great for running local LLMs. That unified memory architecture is genuinely impressive for AI workloads.

But it feels like driving an SUV in an urban environment 🚗. How often do you actually use it at full capacity? Most of the time you're idling at traffic lights. You're much better off renting one when you need it — spinning up a GPU instance in the cloud for the heavy lifting and keeping a lightweight, open machine for your daily work.

Tech sovereignty ⚡

There's another dimension that matters more than benchmarks.

Your Mac calls home. Constantly. Telemetry, iCloud nudges, App Store prompts, subscription upsells for iCloud+, Apple Music, Apple TV+, Apple Arcade, Apple News+, Apple Fitness+. Your $3,000 laptop is also a storefront.

Linux doesn't phone home. There's no corporation nudging you toward a subscription tier. No telemetry you didn't opt into. No App Store tax. No "sign in with your Apple ID to continue."

Your machine. Your data. Your rules.

In an age where AI assistants have deep access to your system, this isn't just philosophical — it's practical. Do you want your AI working on a system that's transparent and auditable, or one that's opaque and phoning home to Cupertino?

The irony 🔄

Twenty years ago, Apple bet that great UX could overcome a hardware deficit. They were right — and it worked so well they became the most valuable company on Earth.

Now the UX advantage has shifted. Not in the traditional sense of prettier buttons or smoother animations. In the sense that matters in 2026: how well can intelligence — human and artificial — operate your system?

Linux is ugly. Linux is fiddly. Linux makes you edit text files and read man pages.

But that's exactly why it wins now. Because AI loves text files. AI loves man pages. AI loves systems that are transparent, scriptable, and open.

The Mac was the "computer for the rest of us." Linux is becoming the computer for the rest of us — plus our AI.

The new productivity gap

Just like in 2003, the faster machine isn't necessarily the more productive one. A developer on Linux with a good AI assistant will run circles around a developer on macOS — not because the hardware is better, but because the system is better suited to how we work now.

The operating system that can be fully understood, fully controlled, and fully automated by AI isn't a nice-to-have. It's a competitive advantage.

Apple taught us that lesson. They just didn't expect it to work against them.

So yes, my Lenovo X1 Carbon can't keep up with my M1 on a benchmark chart. But with Omarchy and a good AI assistant by my side, I've never been more productive. 🌱