The Human Experience | ✉️ #44
Hey! 👋
Leo, Pablo and Kirill are out for today, so please indulge me, a humble editor, in a few paragraphs of AI blabbering, and we’ll get to the good stuff immediately after.
Do you sometimes get tired of talking about AI? Not because it’s out of nowhere or the topic you were talking about suddenly pivoting to AI was unreasonable. Just in general, on a human level, do you sometimes have this feeling that, while interesting, it’s so exhausting just to think about it? No, just me? Anyway, you simply can’t do that anymore, you can’t afford it.
AI is changing our everyday life. Have you checked out mkdev’s podcast recaps already? They are partly AI generated. They could be written by hand by yours truly, but they aren’t, it’s simply just not time-efficient enough anymore. You might think they could also be fully generated. Re-caping texts is one of the things AI is supposed to be really good at after all. But that’s also not true yet, the AI product is still almost unreadable and useless. So the everyday change of my personal life wasn’t “One of my tasks is automated now”, but rather “I now talk to a wall for a few minutes a day and I have a special relationship with this wall, to get what I want”. I didn’t have relationships with software before, that’s new. (Insert a “Her” poster here).
That might be a part of the reason for the frustration AI causes. The disruption it brought is already here, but so many of its potential benefits are still in the future. You might have seen the scandal with MKBHD and a useless pin, or another AI device that could just be an app on your phone, but is a product somebody wants you to buy now. Pablo talks about things like this on mkdev all the time.
The pressure of AI products solving every problem ever while they certainly aren’t at the point where they can do that yet, is huge. But it’s already quite good at some things, so for now we are stuck in this limbo of possibilities and promises. It is exciting for some, but it’s also exhausting. And then we get to the things that AI won’t be able to solve, but that are still affected by it. In a recent essay musician Adam Neely talks about the problem with AI in music. And it’s not another “We shouldn’t generate songs, that’s not art”, I promise. It’s more related to the things from the 35th episode of “DevOps Accents”, about the social nature of many human processes, about the nature of value and other interesting stuff.
AI predictions can sometimes be bleak when we talk about social consequences (here’s another essay, by Kyle Hill, about that). AI is already massive in the scope of its production, making everything a sea of faceless content, a dump of products. But despite a fun quote from a recent “Fallout” show (“You're a product, I'm a product, the end of the world is a product”, says Matt Berry cheerfully), we’re very much not products. That seems to be the heart of frustration in all the AI talk for me. But I’m sure we’ll figure it out, I mean, look at all the promises!
What We've Shared
IBM buys Hashicorp, Pulumi leads the way, AWS embraces Llama 3, Kubernetes 1.30 / mknews 004
Broadcom kills VMware: is this acquisition good or bad for you?
DevOps Accents #36: Google Cloud Next, Google and Spotify Layoffs and Non-Compete Agreements.
Cloud Run without VPC Serverless. Every time we want to communicate with a VPC, we need to use complex elements like Serverless VPC. We have numerous articles, videos, and even a webinar addressing this problem. However, since summer 2023, there has been a beta version that allows us to connect our Cloud Run service or job directly to a component inside a VPC without Serverless VPC.
What We've Discovered
KubeCon EU 2024 Paris: Key Takeaways. A great wrap-up of the most important topics and talks from the latest KubeCon in Paris!
Moving fast breaks things: the importance of a staging environment. If for some (bad) reason you don't have a staging environment, this is a good reminder why it's important to have one.
Post mortem: our outage took 124 seconds from you, here's 378 back. Not a practically applicable, but highly fun to read post-mortem of last-year's 2 minute outage of Graphite.
Building Application Reliability on Top of Infrastructure Unreliability. This rticle has a good point, but it didn't mention any good solutions except using the tool where the article is published. There are so many more ways to improve reliability on your application side and handle failures!
Autonomous hardware diagnostics and recovery at scale. CloudFlare shares how they automatically detect issues in their hardware servers and handle ticketing, maintenance and deactivation of servers without participation of humans during the most part of the process.
A random reminder
Our main social media page is our LinkedIn. Follow us there, if you still haven't!
The 45th mkdev dispatch will arrive on Friday, May 24. See you next time!