[layer8]

> were you insane enough to actually write something useful and publish it you'd just be feeding the LLM's to directly compete against you with your own regurgitated work.

If LLMs need the publication of useful writing to compete with it, then they won’t be competitive without it. In other words, there is always incentive to publish writings that aren’t something you would otherwise be able to get from an LLM.

In addition, LLMs don’t necessarily pick up on a single publication. Their training is more shaped by patterns and concepts espoused by a large multitude of publications. This also incentivizes the publication of novel original thoughts.

The web isn’t dead at all. The web — HN being a great example — is also an entirely different way of content discovery than prompting an LLM.

[apsurd]

How do people find these novel publications?

Seems it's the age of social, 1:1 subscriptions like substack, patreon -- creator economy.

That's fine, but in this ecosystem, there's no need to publish things publicly, right?

Public publishing is for search discoverability. If the end game for search is instant answers, then both the engines and the content creators dont benefit one another. To your point of it only mattering in the aggregate training sense.

So public content discovery by index is dead.

[layer8]

I agree insofar as discovery of novel articles mostly happens through sites like HN for me, or, when looking for specific topics, via tools like hn.algolia.com. But I also use Google many times a day to find stuff, including human-written content, that LLMs are useless for, or at the very least more cumbersome, and/or that I’d have to verify by Google searches anyway.

One thing that the web provides that LLMs don’t is that it lets me form an opinion, by sampling a multitude of online resources, whereas LLMs only give me one take, or a bullet list of takes, without a good affordance on assessing them.

By the GP argument, HN comment threads are dead, because why would you read them (or post a comment!) if you can instead just ask an LLM.

[dageshi]

Those outputting the article from the LLM can concentrate on SEO including backlinks from other LLM generated articles. The question becomes does an average LLM article + the best possible SEO beat the original article?

I'm gonna say, yes, it does at which point there's no incentive for the original author. That's of course assuming the user actually searched at all or went to the LLM directly instead in which case the original author of the article has no chance at all.

[layer8]

People write and publish without such incentives, without the goal of turning up on page 1 of Google search results, and don’t care about being “beaten” by whatever alternative publication. And at least so far, truly good and useful content generally does find its audience, even without SSO.

There’s also a lot of content that can’t be meaningfully replaced by LLMs, such as people writing about their side projects (such as the current top HN submission https://news.ycombinator.com/item?id=41154135), product reviews, or more generally, reports of personal experiences, or personal takes on whatever topic.

[dageshi]

If that were truly the case, this thread would not exist.

The truth is a decade ago the quality of sites you could find and therefore the quality of search results were way higher, back then google was literally magic in terms of finding answers to questions.

It is far, far from that now.

[layer8]

I fully agree that the experience has degraded (though defaulting to verbatim search helps a lot IME). However, I was responding to the assertion that the web is dead and that people will stop publishing on it, which I see little basis for.