I run literally all of that, and a couple of virtual machines on my 2016 rMB. It works fine.

Sure, I can't play Doom 4. But if I need extra cycles for my development work, I spin up an ec2 instance.

As long as the data-sets are test-sized and you're not simulating thousands of simultaneous users, that stuff should be no problem even for very modest computers.

The performance issues come when you're running each of those in it's own virtual machine, and all the overhead and tooling that comes with that.

We run Gitlab (Rails + Unicorn + Sidekiq + Nginx + Postgres + Redis) on a 3GB single-core VPS and it runs fine, without needing the swap. Is the stack really the problem?

Isn't that hugely dependent on the use-case? If you're working with significant volumes of data - which makes sense even in testing to validate that your application actually behaves sanely with data volumes approach real-world scenario - you might not have a lot of fun with 3GB. And optimizing for great on-disk performance during local development doesn't have much benefit if your production workload is all memory.

How many users on your Gitlab? I've had to add 2GB of ram every 3-6 months to my Gitlab server. We're at about 17 users and 300 projects right now. I think it's on a 4vcpu box right now.

8 users, 850 projects, 6GB+ of Git data.

It runs fine, but why not have, it runs "awesome" :)

Why local and not cloud/vps?

Why bother with a "pro" device at all, just get a chromebook and call it a day. Except there's a ton of premium stuff you can do more easily locally, and it's all stuff that hasn't been productized down to cheap web-office quality standards

Many of us have done just that...