The Butt Ugly Weblog

(Posted this on Bluesky, so the story is a bit jagged, as it was originally composed into 9 separate posts.)

I've been using AIs as a programming aide as I am delving into a slightly unfamiliar environment, and I'm realizing I'm not getting as deep of an understanding of the underlying system as before.

I've seen people lament this. "AI creates bad code and programmers don't know how to program anymore."

I don't think it's a bad thing necessarily. We've seen this before when we moved from assembly and low-level languages to higher level languages.

I remember people lamenting how programmers no longer understand CPUs and how they treat CPU and memory like infinite resources. And how owning your own hardware is much better price/performance ratio than going cloud.

Yes. We (engineers) exchanged performance for development speed. We evolved the software engineering discipline to deal with the situation. Most of us now think in components and microservices instead of L2 caches.

Some people still live close to hardware and understand how it works. But it's a specialty now, unlike in the 80s.

So now I have a feeling we're doing the same kind of a jump: We're dumping performance again for efficiency.

Now, compilers for high-level languages like Rust (or even Java) have pretty much caught up with the modern processor architectures and they're not actually that terribly far off from hand-optimized code.

So I think this is going to happen with AIs as well.

Possibly it's going to need more than LLMs, but right now LLMs are like compilers that turn your intention into maybe okay code. We are still badly lacking in the necessary engineering disciplines to deal with all this I think, starting from reproducibility, QA, ops, or even version control.

But I think we can sort of see the light at the end of the tunnel. Power requirements for individual models are going down, and specialized hardware is being born. Millions of engineers are thinking about these things every single day, as opposed to thousands of data scientists two years ago.

Engineers are a powerful, fairly unstoppable force that tends to flow through any crack if it sees a treat on the other side of the wall.

With AI, we see the treat. We just need to see the threats too.

As a teenager, I was a voracious reader. And this continued actually for a pretty long time, but it became more difficult over time. Many have complained about the same thing, and I see a lot of answers: social media, streaming, kids, work, other pressures of daily life... Sure. They all contribute. And I've been saying the same things about my own reading. But I wanted to see if they actually were true for me. So as my summer holiday started this year, I just decided to read as many books as I could and see, if I still can and want to.

I know a lot of people prefer audio books these days, but for me they just don't work. My brain is wired so that I quite often pause to wonder about a passage, or it triggers some memory, or I start daydreaming, or I make notes about it for a role playing campaign. With a book, all I have to do is to lift my eyes from the page, and when my thought is complete, I can just lower my eyes and continue reading from where I left off. Not so with audio books. With audio books, I just lose the passage, the narrative and everything, and while I sort of complete the act of reading/listening to the book, I find that I have missed half of it and have only a generic idea as to what happened in the book.

Having said that, I did listen to one audio book during the summer, which was a pretty cool experience. I went on a walk, then I borrowed the book from the library on my phone, listened to it, and then returned it before I got home. (Edith Södergran's poetry. Poetry is always hard to listen to, since so much depends on the delivery. This was a good audiobook though.)

Aaanyway. During July and August, I was able to read 17 books, ranging from poetry and high fantasy to military fiction and business literature. During holiday, reading was a bit faster, but even after returning to work I was able to read pretty much all I wanted. I started off light, borrowing just one book from the library, and now I am in a state where I have five books checked out from the library and I'm pretty confident I can finish them all within the month. (Yes, one is missing from the pic.)

What I found out was that the reason I had not been able to read as much before as I wanted to was... lack of practice. My skills of reading had simply rusted due to low usage. I started off easy, reading pageturner, not-much-thinking-required books in my native language, just to see when and where I can read. So instead of doomscrolling, I read during breakfast. I read a quarter of an hour here and there. I made the book easily available. I read before bedtime. And my skills returned. I got into more complicated books (Harry Martinson's Aniara especially was quite emotionally heavy) as reading again became effortless like it was during my youth.

This experiment has made me a lot happier and relaxed than before. I've missed reading and books so much.

So I suppose here's a lesson: If you want to read more, give it a try? It might be just rust, and you might not know it until you try. Worked for me; might work for you.

On a colleague's recommendation, I ended up grabbing The Phoenix Project from my local library. Which has, pretty much since COVID, become my happy place. Not that it wasn't before, but the biggest thing I found myself missing during lockdowns was that I couldn't go to the library. Everything else I could cope with, but not being to access a library at my whim felt really, really bad. So I now make a point to visit it regularly.

Anyhoo, this was a weird book. It is a hero tale of an IT manager, who suddenly gets promoted to a VP of IT operations and is suddenly responsible for saving a multi-billion company from serious trouble and his entire IT team from being outsourced. And naturally, he accomplishes this with the help of a mysterious stranger who brings in his sensei-like wisdom about DevOps, and a few people who are eager to grasp onto these pearls of wisdom escaping his lips.

As a book, it's shit. Characters are uninteresting, plot is thinner than the paper the book was printed on, and I could not help but facepalm occasionally. There are three kinds of people in this book: Smart and Wrong, Stupid and Wrong, and Smart and Right. Of course the people who are Smart and Wrong become people who are Smart and Right, because of DevOps wisdom sprinkled by the protagonist and his sensei. And people who are Stupid and Wrong are the main antagonists here, and they of course lose. Because the hero story, right?

But as an illustrated case study for DevOps principles, it is brilliant. Basically this is a list of all different issues your company might be having, ranging from too much work-in-progress to managing unplanned work to managing conflicting business priorities and whatnot. It's easy to put yourself in the shoes of the protagonist, and just execute smart plans using solid principles to solve each problem as they come along. In that sense, it's not that different from a lot of hero epics, such as The Martian by Andy Weir (which, unlike this, is actually a good book that I recommend).

This book is an interesting approach. It has certain, well, aynrandistic preachiness in it, but still a free-flowing fiction might be in some cases a good way to make your reader to think about the ideas you have. So if you read it as a marketing text on DevOps, it's pretty good. It's easy and fast to read, doesn't really slow down anywhere, and comes with a useful appendix of an excerpt of the business book they really want you to buy after this one.

Just don't expect a great narrative, well-rounded characters, or actually, well, caring about any of them.

Just a bit of ye goode olde haircut blogging, just because I feel like it today. I mean, I could write this up also in Facebook, but I'm getting a bit tired of it, really.

But hey, so I wake up this glorious Monday morning, wondering why my alarm clock didn't go off? But hey, I am only a few minutes late, and I have plenty of time in the mornings usually. I like my morning routines, waking up slowly, reading the news, etc.

So the news isn't obviously good (because it seems that there are two nuclear-capable nations just lobbing missiles at each other), so I get distracted a bit. As I am slowly moving towards leaving for work, I check my bank app. It's payday, so I sortakinda want to just see the numbers, as it's also "pay my monthly house loan dues" -day.

Nothing. No money in, no money out.

WTF?

Slowly it dawns on me. Is it... Sunday? Could it be?

Oh darn. It is Sunday. But hey, that's good right! I'm already up early and ready to go!

(A few hours later I've squandered most of the morning by debugging some spam filter issues and ugprading things left and right...)

Although... Why didn't I notice it's Sunday? Probably because in the digital world, Sunday is pretty much the same as all other days. No more Sunday issues of newspapers - they all look the same in the app. No morning TV - Youtube and Netflix have the same stuff every day. No morning radio - it's all just Spotify. So it's a bit weird to realize that one thing that this hyperpersonalized age and information work completely deletes is the gut feeling for passage of time. In a way, it is a return to earlier times, when seasons where the defining factor in people's lives. Industrial age was very much about dividing labour into hours and weeks and months over seasons and years, but information work does not care. We expect to have our services available when we want them, not 9-16 on weekdays. Information workers are not only allowed, but sometimes encouraged to ignore "regular hours". And the further we drift from the industrial age, a day, week or month is pretty much the same as any other. Seasons we notice, years we feel passing.

I take no value judgement on whether this is good or not. But the change is likely not to halt here.

Well, I'm moving this server physically from Germany to Finland. Or to be precise, I rented new hardware, moved all my stuff, and am now in the process of rebuilding everything on the new server. So if you're reading this, at least the web servers have been moved properly.

Anyway, moving digital assets is much the same as moving in the physical world. It always takes longer than you plan, you invariably break stuff, and you are constantly amazed at the amount of cruft you have accumulated over the years and you SWEAR you're going to clean everything up AS SOON AS everything has settled, and then you basically end up putting stuff in a big box that says "TODO" and go open a beer. And when the next move comes, you open the box and you go "Oops, do I still need these?"

There's another surprising thing that is similar, and it's that configurations don't move well. It's very difficult to end up with the same order of books on the shelf in the new place, or the same configuration in the kitchen, and most likely you don't even want to. Some of the stuff you just chuck in place, and for some you spend hours contemplating what would be the best configuration be - should the sofa be in that corner? What's the best way to get lighting on my desk? Should this lamp be there?

Same with digital stuff: Web server needs to be reconfigured; your <program> config needs to change (since the original is probably from 2012 and the server config file format has changed)... When you own your furniture/server, you need to think about configuration as well. If you have a web hotel and just copy stuff from one place to another, it's almost like moving from one furnished apartment to another - no need to think, just fill the cupboards and you're done. But when you insist on renting just the walls/server, you need to configure things on your own.

And I know I don't need to. Most of the things I use my server for is, well, you could just rent the equivalent - probably better even - from a SaaS provider. And modern software development revolves around docker containers and serverless these days, so there's little to no need even professionally to know the ins and outs of servers anymore. And this tinkering can be a bit frustrating too. However, I do appreciate the fact that by tinkering on this server I gain far deeper knowledge of how the internet really works on a very concrete, even visceral, level. After all, I've always been a bead curtain kind of guy.

---

Oh, and Happy New Year 2024! I think this blog can now legally buy beer even in the US...

So I tricked my brain by reserving an appointment for today at 9:10, because I know that stupid little pile of noodles will round it down to 9:00, and I won't be late. Even if the road conditions were bad, like they turned out to be.

So does this make me very smart or very stupid? 😜

What slightly bothers me about a lot of the AI coverage is the discrepancy between the hype and reality. Some folks are screaming AI apocalypse, and some folks go and "I tried ChatGPT and it gave me shit answers."

Just a few years ago, we had an equivalent situation with EVs. Some folks were saying that Tesla will build all cars in the future, and others just saw a big bunch of panel gaps and range issues. That they would never replace diesel. And you couldn't charge them everywhere.

Yet almost overnight, EVs and chargers are suddenly everywhere (in richer countries). Tesla is big, yes, but not everything [in Europe & China; US is weird]. It just took the understanding on how to make them cheaper and faster to build, and then building the factories and supply chains. Fundamentally they are better vehicles, but the first implementations were buggy as hell. They are not perfect, but they have become _good enough_.

There's a ton of panel gaps and quality issues in current generative AI tools as well. But consider the speed at which they are improving - ChatGPT is nary a year old - and we don't yet know if they are hitting any sorts of limits. It again becomes a question on how to make them cheaper and better so that they can become ubiquitous. We're still pretty much at the "throwing spaghetti at the wall and seeing what sticks" -phase of AI; figuring out where and how it could and should be used.

Neither do we know what the real dangers of AI are. EVs didn't become a public menace because they are so quiet either, even though that was a big argument against them a few years back. We solved it by making all EVs make a mandatory noise.

So don't take the current AIs as representative of the future. Neither take the hyperbole. Reality, as always, is somewhere in the middle.

(Unless someone gives them weapons and makes them eat organic material and self-replicate. Then we might have some Horizon Zero Dawn -level problems.)