Hardware hasn't been commoditized

I remember when you could pick up an AWS micro instance for a couple of dollars per month. It was a no-brainer. To age myself a bit, I believe the first thing I used such an instance for was to host a Turntable.fm bot so that my co-worker and I could host our own channel during work.

It was a couple of dollars back then, so it could only get cheaper for equally-powerful instances going forward right? Hosting in the cloud was always going to be cheaper because of economies of scales and efficiencies and whatnot. The future is one in which the operating system has been abstracted away and people only worry about what they run not where they run it or what they run it on.. right?

What a joke that has turned out to be. Either that or I've long since entered the grumpy old man phase of my career. But no, I think my gripe here has merit. The modern-day version of my Turntable bot is this blog, or a NextCloud instance that I run to sync my files between devices. Neither of these services need much at all in the way of processing power or memory, I can mostly pick the cheapest instance on the cheapest platform.

But once you add in a bit of disk space, or even a couple of GB or RAM you're in the $8-$10/month range already. If you want some breathing room for either memory or CPU you're looking at around $15/month. So $15/month for a couple of cores and a 2 GB of RAM? A last-generation Raspberry Pi 4 with twice as many cores and 2GB of RAM is around $35 and you own it outright.

I get that you're getting some other benefits beyond the stats: bandwidth is probably better, it's probably more secure, and you don't need to install the OS yourself etc. But for anything that's not a business-critical production thing I'd be extremely hesitant to spend more than I do on Netflix for hardware that's half as fast as a Pi and I don't own anything at the end of it.

No, hardware has not been commoditized. The breathless promise of the cloud hasn't panned out from my (admittedly very limited and very niche) perspective. Turns out maintaining infrastructure is tricky and expensive. If you want someone else to do it, you're going to pay. 37 Signals, the owners of Basecamp famously left the cloud recently. I know there's no real comparison between my hobby projects and their enterprise systems, but it's funny to see that it happened around the same time that I became frustrated with my monthly blog costs.

Their post is very much worth a read even if you don't work in the space. I absolutely love that their willingness to go against "industry best practices". They were lambasted for their move from the cloud just as they were recently lambasted for their move from Typescript. I feel a real affinity for DHH, he's controversial and opinionated but he's sure made several moves that strike me as common sense in the software space.

One strange theme for me as I've advanced in my career (supposedly to "senior" developer status) has been becoming way less sure of myself rather than more sure. The more I learn the more uncertain I am. This is true for my own decisions and actions, but it's also true for interactions I have with others. In spite of myself, when someone speaks in absolutes, giant, blaring alarm bells start screaming in my brain and I want nothing more than to run away as fast as I can. Oh, you think this is obviously "the answer"..

I'm probably insufferable as a co-worker and human as a result of this foible, but evidence keeps piling up that uncertainty is a decent default in many situations.

So, back to this blog. Both the blog and my Nextcloud instance are happily running away on two old Raspberry Pis now and have been for many months. They're faster than the DigitalOcean instance they were running on by orders of magnitude and sip power and I'm very happy with them.

I'm not ready to accuse either AWS or Digital Ocean of engaging in the enshittification of their products, but I'm still bitter about both the IT hype cycle generally and the cloud-gasm specifically. As a result I'm not willing to take your course on AI and I don't want to be certified in your new thing.

Is that a blind spot of mine, or well-earned wisdom? Is it the privilege of someone who has experience and can dismiss trends and still get a job? I guess I can't know right now.

But I see a class of punditry in IT that mirrors the general class of punditry in other domains and utilizes the same tools and relies on the same tricks of psychology–that being that we reward and remember the soothsayer who predicted the thing more than we punish the person who predicted the thing that didn't happen. Because of that, the marginal cost of spewing bullshit is relatively low compared to possible gain of hitting the bullseye with your random-ass buckshot.