Wut?

This is not an extraordinary claim if you’ve used a good LLM in the last, say, year.

Yes - it’s still useful to have docs to refer to.

But it can also be useful to use the AI to generate that documentation if you’re not good at it (which I am not) or aren’t disciplined enough to keep them up-to-date (which practically nobody is).

I think he’s maybe overblowing the impact of this change.

Maybe… But I don’t look forward to working with the dozens of IT folks at my company who all installed git once and have never upgraded. This will be multiple meetings, show up on slide decks, break builds…

It’s going to be a right pain for nothing.

Maybe - but it’s unrealistic to expect. I tutored my high school nephew in a programming class he was taking and practically all of his classmates were using AI to do their homework assignments. The genie is out of that bottle.

College students should at least be taught how to use it and have AIs provided that help them to learn. You can make an AI that helps with the “what questions should you be asking” at least to help them with that insight they haven’t developed yet.

For example - I’d recently read about an AI that was created to help grad students with their research proposals. Not in doing the proposal per se but helping them to think about their proposal and whether it was achievable, testable, adds to the corpus of knowledge, etc.

AI can answer many of their questions, but AI can’t help them with the questions they don’t know to ask.

This is a very important thing to understand and one that is difficult to explain to non-technical folks who have a new “super power” where they can magic-up programs (and why hosting their vibe-coded monstrosity on our servers and network may be a bad idea).

This is why when we say “AI is a tool” we need to be careful to add that it’s more like “heavy machinery” than a “screwdriver.” You need to be checked out on it and understand what it will be doing for you and how to work with it. It’s not a tool everyone is readily equipped to work with.

We seem to have forgotten that humans create a lot of low-quality and low-effort content…

Over at sh.itjust.works we recently had a contest to replace our previously AI generated logo (because “AI is bad now”). The human-created replacements were a parade of low-effort copy/paste garbage that were embarrassing to judge… But they were human garbage so we picked one.

If you don’t like it it’s AI slop.

Maybe I misunderstood - The “NixOS (bare metal)” seemed to imply to me that the containers were not bare-metal by omission. If not then my complaint is retracted.

I’ll die on this hill.

Containers run on “bare metal” in the same way that other processes on your system run.

Let’s assume I have a software application with millions of files.

rsync is generally good at things like this.

I would just be creating records within the main database.

So is it one file? Or millions?

I really wish kscript would catch on. Kotlin is an excellent language and I have enjoyed using it for some utility scripts.

Kscript catches dependencies and compiled output. First run compiles, second is fast.

I thought it was clear that I was responding to “if you forget to assign a value to one of them, since the object constructor will likely assign an unknown default value.”

I’m not defending the article.

That depends entirely on how the object is constructed. Plenty of implementations use object.new() and you’re expected to fill in the blanks.

That’s because some languages simply do not give a shit about data types. They’re just like “your opinion man” and can even vary at runtime which creates lots of exciting errors as you’re pointing out. Enjoy writing lots of type-checks.

In a language that does care about types you can control that and don’t worry about it.

Besides, even if you have a constructor that takes parameters, it’s essentially the same as calling a function. Get it wrong and stuff breaks.

data class UserRequest { 
   val id: Int,
   val name: String
}

No guesswork. All values must be provided and none can be ‘null’. Well designed languages help you avoid such problems. Poorly designed languages facilitate it.

if you forget to assign a value to one of them, since the object constructor will likely assign an unknown default value.

This bit - in a language that cares about types that should either not be possible or difficult to do by accident.

If you’re using a language that cares about data types that’s not possible.

AI tends to “discover” previously discovered vulnerabilities,

Friend - your information is very out-of-date. It’s accelerating the discovery of vulnerabilities like crazy. I’m not saying we’ve hit singularity (or even that we will) but to brush it all off as “it only finds things we already knew about” is copium at this point.

The SciActive Human Contribution Policy is a free, general contribution policy meant to safeguard projects from the inherent legal and security risks of AI generated contributions.

Boy it’s amazing how people ignore the security track record of human authored code over the last 30 years. One need only look at the recent explosion of discovered vulnerabilities by AI.

This license is fud.

Oh, wow, I never considered that art could be created from code. You totally got me.

Or not - the fact that you had to modify the code to turn it into art proves my point.