Up front, because it belongs to being honest: I am not paid for this. No affiliate link, no sponsorship, no discount code, no commission. Apple doesn’t know this text exists.

And really I’m the last person who should be writing this. I grew up in a pure Windows household, my father a Windows man through and through. I started building my own PCs early, gaming rigs later; my brother is still at it today. Mostly because of video editing it was clear to me for a long time: PC, that’s where the power is. I bought fat towers and I was convinced.

Today my whole studio runs on a MacBook. The commercials, this site, the shop, the apps, this newsletter — all of it is made on a Mac next to an iPhone. How that happened, and why I don’t go back, is the most honest way to explain the thing to you. The “where it’s annoying” section is in here anyway — a recommendation without the flip side isn’t one.

From “gimmick” to the only tool

In my first semester of university I still had a Windows laptop plus an iPad for taking notes. All around me sat classmates with MacBooks, and I honestly thought: nice gimmick, doesn’t actually do anything. Until I had it in my own hands more and more at friends’ places and slowly understood what it’s really about — not the case, but the operating system.

The reputation that Apple is “limited” is true on the surface and false underneath. If you really push it, macOS is a tool for developers and creators — Apple isn’t one of the most valuable companies in the world for nothing. I went over once and never went back. And I’ll stay, as long as they don’t completely blow it: this ecosystem is built for designers, for creators, for people who want to realize their ideas fast and from anywhere, without the device getting in the way.

The reason that tips it for AI: one memory for everything

The most important difference is invisible and sounds technical, but it changes everything: unified memory — a shared memory pool.

On a classic PC, the graphics card has its own, hard-limited memory. If an AI model doesn’t fit in it, it either crawls or won’t run at all. On Apple Silicon — Apple’s own chip line, currently the M5 family (in Air and Pro) — the processor, graphics and the AI engine share one memory pool. The system memory is available to the graphics and the AI as well, not just a small, separate pool of graphics memory.

In practice that means: a Mac with plenty of memory can run AI models locally, on your own device, that a pricey graphics card with less of its own memory can’t even fit — and without the hard drop-off that hits normal PCs the moment a model no longer fits in the graphics memory. (Honestly said: the Mac doesn’t reach the raw speed of a top-end graphics card here — its advantage is that the big model fits at all.)

If you already have an iPhone, it clicks in

The second argument doesn’t apply to everyone, but to most people here: if you use an iPhone anyway, the Mac works with it as if they were one device. A photo of a sketch is on the computer in seconds, no cable. AirDrop fifty things over quickly. Text you copy on the phone, you paste on the Mac. Calls, messages, the same passwords, the same iCloud data — when you’re building, it’s all connected. It isn’t one big feature; it’s a hundred small points of friction that simply aren’t there. You only notice it once you’ve had it for a while — and then a lot.

At its core it’s a Unix machine with a nice lid

For anyone who builds things — programming, websites, automation, exactly the things from my Claude Code post — the most important point is beneath the surface: macOS is a Unix system. The terminal that developers worldwide work in is there out of the box and behaves the way it should. Installing tools, running scripts, operating an AI assistant in the terminal — none of it is a fight against the operating system.

An example from my day: I take a screenshot in seconds via Spotlight, drag it into a folder, hand it to an AI, have it transformed and drag the result back — that flow is seamless on the Mac. And tools like Claude Code ran directly here from day one, so you could try a lot early. The pretty surface for everyday use and the powerful command line for the work are the same device, no compromise.

Quiet, cool, all day

Because Apple designs the chips itself, they’re extremely efficient — and in daily use that’s very concrete. I have my MacBooks on basically all day, every day. In years none has crashed on me from having too much open, none has overheated, none has stuttered; the build quality is simply there, and the device just runs quietly. The fan rarely spins up, the MacBook Air has none at all. The battery lasts through a workday without you eyeing the outlet. For working on the move — on the train, in a café, on set — that isn’t luxury, it’s the difference between working and hunting for a cable.

The move that excites me right now: a cheap MacBook + AI

Apple made a smart play. Instead of forcing its own closed AI — they spared themselves the “Copilot move” — they bet on their own, extremely optimized chips and stay more open to integrating other AI tools. And because phones are by far the biggest demand, the enormous optimization of those chips pays off — the same chip family that’s in the current iPhones and heavily tuned for AI can then be put into a computer cheaply.

That’s exactly the new, cheap MacBook Neo: the chip from the current iPhone, plenty of AI power, a small price. To me that’s the most exciting Apple move in a long time — and the practical point for you: A cheap MacBook plus a Claude subscription is, today, something like a complete knowledge team on your desk. It runs from anywhere, you can drive much of it from your phone, and with the automations from the Claude Code post a single person builds things out of it that used to take a team. That’s the thread behind almost all the AI posts here.

Where it’s annoying — honestly

Now the flip side, without which the recommendation would be worthless:

  • The fast models are expensive. At the top end you pay noticeably more than for a comparably fast Windows machine — part of that is substance (chip, display, build), part is simply brand. At the bottom Apple opened up the field with the Neo, so the blanket “a Mac always costs more” rule no longer holds.
  • Memory and storage are soldered in. You can upgrade nothing later — what you buy is what you have forever. That makes the decision at purchase important (see below), and Apple’s surcharges for more memory are unpleasantly steep.
  • Repairs are restricted and expensive. A lot is glued, and some repairs require a software calibration step — which you can at least now run yourself via Apple’s Self Service Repair. For a device meant to last, it stays irritating — even though Macs tend to hold up well and resell well.
  • It’s a closed system. You do things Apple’s way. Anyone who likes to get deep into the internals, tinker with the hardware or wants maximum freedom is happier on a Linux or Windows machine.

What I’d actually recommend

No prices, because they age — just the logic:

  • To dip in, and for an AI setup on a small budget: the MacBook Neo. That’s exactly what it’s made for — cheap, frugal, with an AI-strong iPhone chip. For many, the best first Mac there is.
  • For most “real” work machines, the MacBook Air is enough. Silent, light, fast enough for everyday use, writing, the web, solid photo and video editing.
  • Reach for the MacBook Pro (with M5 Pro or M5 Max) only if you truly need it — heavy video editing, large AI models locally, hours of full load. For office and web it’s overkill.
  • The one tip that’s worth real money: don’t skimp on memory. Because you can’t upgrade anything, it’s the one decision you’ll regret later. Better one step more memory than one step faster a chip — especially if AI is meant to run locally.

I’m no Apple fanboy — I come from the self-built gaming PC. But for the way I work today — solo, mobile, lots of AI, an iPhone in my pocket — it’s the tool that gets in my way the least. And that’s what a tool is for: to let you make things, instead of keeping you busy.

Sources and status

Model and chip details are as of July 2026 (current line: Apple’s M5 MacBook Pro announcement); the chip generations change yearly, the arguments underneath don’t. On repairability, the critical counter-view from iFixit is worth a look. This site contains no affiliate links and is in no way supported by Apple.

Questions, or disagree? Write to me.