WRT union comment, that has no relevance to quality control. Actually there's very limited evidence to the contrary in other industries. Take cars for example: All of the models that UAW lists as the ones they manufacture score very poorly on reliability in consumer reports. That could also be poor engineering, but either way, it's unwise of them to claim that they produce quality cars when everyone but them says the opposite.

Software in particular -- compare software produced in the US against other countries, where developers in those countries are often union. It isn't even a contest.

Either way, I don't see any compelling real-world or practical reason to believe that unions will help your argument.

Nor did they choose what libraries or other tools you used.

Evidently not. You're literally making a big deal about what tools were used without any mention of whether they're used correctly.

Go tell that to the C++ developers who refuse to care about memory safety. Pretty much every last one of them thinks they're incapable of creating undefined behavior, including the guy who created the language. These are bad engineers. You don't trust bad engineers. People who drive cars made by Volkswagen and/or Stellantis do so because bad engineering doesn't bother them.

Anyhow, rust is my daily driver language. I use it for everything unless there a compelling reason to use a different tool. Rust is a tool with many guardrails, so it's harder to misuse it. But just because a newer circular saw has a thumb guard doesn't mean you can't cut off your thumb with it.

I don't think so. LLMs are just another tool in your belt. You're talking as if misusing the tool is the tool's fault. Not vetting its outputs is definitely misusing it.

You don't, and it's foolish to assume that the use of LLMs makes any difference here. And that's exactly why you don't just install every program you find on the internet just because it's given away for free. If you do, then you're misusing your own tool, namely your computer. You either trust the developer, or you don't, or you do your own vetting by running it in a sandbox to observe its first.

That doesn't change the fact that it's always on you, as an engineer, to know what your code does. Especially in the age of supply chain attacks, like the recent openssh backdoored from a dependency. That didn't even involve AI, and there was a lot more at stake than a stupid credit card number.

In fact, a practice I commonly do before deploying a release is to look over the manifest (i.e. cargo.toml, pyproject.toml, gradle, etc) and see if there's anything I don't need anymore, and ensure that the selected version of all imported libraries are at least a month old, unless one has a badly needed feature or bug fix.

If you're not vetting your own code, then you're a bad engineer and you should really question your life decisions. Period.

I haven't tried to vibe code with any publicly available (that I'm aware of, but it changes often) models in a while, but in my experience, python is itself slop, and has mostly always been that way. Whenever I use it, I strictly use pydantic and enforce mypy and turn on literally every single ruff rule, only later tuning and disabling the ones I don't need for that project as I'm writing it. That highly limits the amount of slop by any maintainer, human or llm, or even me on occasion when my prototyping carries on too long.

Mind you, python has its uses so I'm not dogging the language outright, just pointing out the fact that it lets you do things that shouldn't be done in any language, and I rely on those tools to maintain my own sanity. As far as LLMs go for python, I think it's a case of garbage in, garbage out.

Without putting opus 5 to the test, I'm inclined to say that it can't build code like that using the enforcement mechanisms I impose, but python being python, there might be ways around it. I highly doubt it could do that in rust with all of clippy enabled and enforcing snafu, no panic and no unsafe. Without those, I've seen it do odd shit, like unsafe calls into ffi when a native, platform agnostic crate is already available. Without those enforcement mechanisms in place (which I've always used,) it probably would produce crap, even if the crap actually works as intended.

One of these days I might experiment with that -- do the same prompts with controls as I do without, then observe the difference.

If you release code, it's on you to know what's in it and what it does, doesn't matter who (or in this case: what) wrote it, contributed, etc.

I think the scenario you're describing is more likely to occur with some rogue library you've imported than it would with an LLM, specifically because it's unlikely that you've read all of the code in every library you've ever imported.

When they do that, I start micromanaging its styling in various ways, i.e. giving it a hard cap on the number of lines in each file, telling it to organize modules in a way that makes sense and keeps like code together, use subdirectories, keep data structures and functions separated, etc. Works best with strongly typed languages. Works especially well with rust given it can arrange everything freely without having to worry about cyclical imports, in addition to the borrow checker and result types adding their own natural styling constraints. Often I'll just be writing my own code, and if I decide to do a big refactor, I'll whip out the AI since it can knock it out really fast and generally does a good job of it.

It obviously hasn't occurred to them that these companies aren't profitable.

Instead of a walled garden, apple is more like a movie theater with airport grade screening to make sure you can't bring in your own snacks. And when the Usher Agent gropes you, he says it's for safety because outside food might be past its use-by date. Then the snack bar only has a choice of two different candy bars, one costs $999, the other that adds peanuts is $1,799.

About 90% of it. In the sense of micronutrient and macronutrient content, it's virtually indistinguishable from lean meat. That includes its b complex and creatine, the kind of things that keep your brain healthy and energized, and are commonly deficient in vegans.

Honestly I don't understand you guys. Basically you're offended that people would eat more of the cow, making its usage more ecologically efficient, while at the same time griping about ecological efficiency of eating meat. As far as I can tell, you just don't like it because of its physical appearance before it ends up on the plate. Which is especially insane given the way you guys resort to highly processing your food just to make it nutritionally viable and somewhat palatable, where it often has a much worse appearance (and smells like dog shit) before reaching your plate than "pink slime" does. Not to mention the way you have to source ingredients from all over the world just to somewhat meet your micro needs, which is ecologically inefficient, and still doesn't even give you what's already available locally from animal products.

At the very least, pick a fucking argument that makes fucking sense. And no, you're not going to convince me to eat nutritionally deficient vegan crap food that's so heavily salted and/or sugary just because it has to hide the fact that it doesn't have any flavor. The worst among you are the ones who demand organic vegan. Remember the b complex and creatine I mentioned earlier? To get a full complement, you're either synthesizing it or you're getting it from animals. Period. Your body and your ideology obviously hate each other. Such a first world problem.

If it's anything like the Oscars, the whole thing is determined by whoever pays the highest bribe to promote their crap. I don't get why people get excited over the world's longest, most boring advertisement.

And Mel Brooks is 100. Age doesn't make them any less relevant, the problem is you're just a fucking prude. Guess what else was popular among the youth in their times? Temperance, fascism, and Maoism. Being an asshole knows no age minimum or maximum.

These guys, though old, age well. Prudish assholes like you do not. History will look favorably upon them, where you'll go down more like Kellogg or Graham.

First, don't invke a "both sides" fallacy. It's stupid.
Second, is there any evidence anywhere that the US engages in any of these? Because it's already known that Russia does both, and in fact uses chemical weapons such as vx/novochik
Third, president Nixon issued an executive order mandating that no government entities engage in this in any way at all, and unlike your tsar's decrees, executive orders have no effect if they're not made pubic, so they can't be secretly cancelled, amended, or ignored. We have a system of checks and balances, you do not.

So first it was opt in, but because nobody really wanted or asked for it, it became opt out. Still, nobody wanted it, so now they don't even get a choice. And ifans wonder why I don't like apple.

Probably worth mentioning at this time that Russia never stopped its bioweapons (and chemical weapons) research and development, which includes inducing antibiotics resistance into deadly microbes.

If this is what that is, you already know you can count on incompetent Russia to hit itself with its own weapon.