Witches can be found everywhere once you start hunting for them.

And even if it was 100% vibe coded, how could one tell? The code has never been published before so there's no way to determine its origin.

Pride and Prejudice and Zombies is based off of public domain sources but it is nevertheless under copyright itself. So even if the output of LLMs was public domain (not something that has been clearly or universally established) that doesn't mean a project incorporating it would have to be. Public domain is not "viral."

You have an idiosyncratic definition of "extreme" then. And "cluster" for that matter.

You guess wrong. C++/C# applications, with a fair bit of Python for various supplementary tasks.

The implementation. I like coming up with the ideas, grinding out the code to make those ideas actually happen is tedious.

I don't know what you would call extreme, then. There's literally only one country on the chart that's less excited about AI, Belgium.

reading tech news left me so confused. all these people around me were letting the language models spit out crap code. they gave up on what i thought was the most interesting part of the job

I think I see the source of your confusion: you're assuming that everybody has the same attitude towards things that you do.

I find that LLMs remove all the most boring parts of the job.

Depends where you look, "the Internet" isn't one unified place. Social media forms bubbles easily, and social media like Reddit or the Fediverse is practically designed that way - minority voices get downvoted, blocked, banned, and so forth. So unless a forum is taking significant effort specifically to ensure diversity of opinion you're going to end up with things drifting to some sort of extreme.

We shouldn't even be in this situation, where just politely asking someone's computer to delete files is effective.

I'm doubting we are in this situation. From the article:

Elsewhere, the Java developer said that Anthropic’s Claude AI code tool flagged the malicious instruction without following it.

The "disregard previous instructions" trick is really old and has been trained for by modern LLMs and accounted for by the structure of modern agent prompts. LLMs can be given blocks of text with a framework that makes it clear thar the text is just data to read, not instructions to follow.

I expect this will be like Nightshade was for image AI - something that anti-AI users degrade their products with and feel smug about but in the end only harm themselves with.

I set up an image generation AI on my computer and it just sat there not generating pictures of anything, until I told it what I wanted it to generate. Then it generated that.

Well, kinda. But the internet is global and you can hook a data center up to it anywhere. There's plans to build them literally in space. And I run local models on my own machine at home, that's not going away either. So the impact of regulation will be limited and local. AI as a whole isn't going away.