The Rise of Whatever

https://eev.ee/blog/2025/07/03/the-rise-of-whatever/

I share what I think is their fundamental viewpoint of: we should care about what it is we're doing here (gestures broadly) and try to make it better

I struggle to continue whenever I read somebody say something like this:

It fundamentally doesn’t do the thing that its investors and diehard fans say it does. It just strings together text that is statistically plausible.

This sort of argument was compelling to me before I had ever used an LLM to write code, but with some practice there's no doubt to me that having something that has digested the web and is able to put it back together with a natural language interface is extremely useful to me.

While it's true on some level that that is what it's doing, it's the same sort of reductive argument as "why does anybody play basketball, it's just throwing a ball in a basket". Which is also true in some sense, and sometimes it's funny to try and view it like that, but it's also obviously more false than it is true.

This doesn't mean that there can't and aren't principled anti-LLM positions, and even that I find some of the arguments in the article persuasive, just that I struggle with that particular form and many like it.

It took me some practice to get used to using LLMs, and now I find them very useful and I'm becoming more aware of the corner cases when using them. I have serious concerns about the ethics and morality of using them, but I've chosen to give them a try.

I'm not opposed to refusing to use them on the basis that they prop up a system you don't want to support, that they're a tool for control by tech oligarchs, that it's unethical to gather and remix the entire content of the internet and librosphere. If you don't like the vibes and don't want to use them, that's great, go ahead!

But it seems like the author is not willing to believe that any sensible person can say in good faith that they find LLMs useful, and my experience is contrary to that.

I like programming as well! I find myself empowered by the LLM, especially by its ability to help me get unstuck.

I worry very much about the people that control it, the structures it enables, and what it will do to our information commons. But I also have found value in it - it has helped me to do things and make things, as the author suggests.