Linux

I recently got assigned to a project where portions of the UI has been LLM vomited. It does not conform to web standards, and it doesn’t give a fuck about accessibility.

There’s a custom checkbox component whose label isn’t clickable. I’ve had to fix so many little things because someone couldn’t be bothered to do it right and chose to outsource their thinking to something that can’t think.

My view on LLM usage is essentially, if you don’t give a fuck about what you’re doing, why should anyone else? If you don’t care enough about your project to actually develop it yourself, why should Flathub platform it?

So this is some what problematic. Technology has benefits does not mean it has to be allowed. Question is what do we gain and at what cost. If the answer is we gain speed, then the followup is "does it really matter?".

Speed of coding is the common answer. Another one is non programmers are able to make programmes. Both of it does not add much of a real value is the point. Because when the machine churn out code at a very high rate, then the humans who should review it gets under pressure. We are not addressing the real bottleneck, instead we just pump out more material which just intensify the congestion at the bottleneck. Quality goes down, we get things like AWS going down, Microsoft wiping people's machines, etc. In other words it creates problem without much meaningful gain.

Regarding the non-programmers making programmes, it mostly just create a lot of noise. There is a lot of weekend projects, one shot attempts, or some half cooked outcomes. Only a handful of projects with meaningful impact in the world will be born out of it. We all know, even before AI, out of hundreds of thousands of GitHub repos only a small section is actually well maintained, and deliver something to the world.

I don't have to explain at what cost. It is just literally affecting the climate and accelarting the collapse of current ecosystem. Are we doing this so that a code can be written in 1 hour, instead of one day? If you get more free time because of this, well, at least we can say the load is reduced. But we all know we just get more work to do from our employer. So I don't really think this is a net positive.

What if I’m in-office dev?

That is an absolutely terrible standard. In fact, that's technically not even a standard. The "I know it when I see it" measure is literally the logic used in censorship. It allows cognitive biases to seep in with no check. A lack of hard metrics means that there' zero ways there can be any objective consistency. And finally, this kind of rationale makes the biggest sin that I can think of, "non-falsifiability".

Whatever people's opinion on AI are or are not. This logic should wholly be rejected in every instance it is brought forth. It is literally the antithesis of rational thought.

I agree that you can tell if the whole code base is AI written. But stuff like dependency upgrades being handled by an AI is impossible to detect, IMHO.

That reminds me of my friend who said (a long time ago) that Japanese girls are prettier than Chinese girls, I asked him how he recognises them, since I knew he didn't know any Chinese or Japanese personally, he said, that when he sees an Asian girl, he just knows, if he likes her, she is Japanese , if he doesn't like her she must be Chinese (and no, he didn't talk to any of them to confirm his "genius" theory... Hey, we were both young, talking to girls was difficult, even more difficult than realising there are a lot more Asian countries than just two)

What I meant is: it's stupid, if you recognise AI assisted code because it's bad, maybe just say that you don't accept shitty code

I'm no competent, or no professional, dev, yes, but I absolutely cannot tell if it's AI generated, let alone AI assisted(whatever that is) code.

That only was the case with ChatGPT3.x era.

Pointless exaggerations aren't inevitable either, but you adopted and shared one. Pointless and "witty" exaggerations won't change the outcome of the process though. Which is why robust and transparent AI policies are required.