Ask HN: Is it time to fork HN into AI/LLM and "Everything else/other?"

Hacker news, like everything in tech, is susceptible to hype. Today it's AI, a few years ago it was Bitcoin.

I do think it's worthwhile to occasionally have a discussion about what content we want to see, and if a particular topic is getting too much attention.

It's also totally reasonable for a group of people to not want their agenda hijacked.

So, IMO, let the discussion continue. Let's see what comes out of it.

To begin with, this would be a non issue if HN just introduced something like user provided tags and users can vote for/against (to circumvent abuse)

Then the people wanting to filter "x" could just do it via simple grease monkey scripts or if HN natively supported it.

Sure, it wouldn't be perfect, but neither does it have to be.

> To me this seems like the subtext is "Those people are not welcome here, they are not like us. It's not like we have anything against them, we just don't want them ramming it down our throats"

It could just as easily be "I don't feel like there is a place here for me anymore and I wish I had another place to go"

> I have seen this question asked on subreddits, "Should we divide this group into A and B, A stays here and B goes over there and that way everybody is happy" To me this seems like the subtext is "Those people are not welcome here"

I don't disagree with this observation about Reddit. However, I feel HN readers are more topic-oriented. Folks really do come to HN to read the articles and then maybe get drawn into a discussion.

I grant there are some topics here that tend to be more engagement driven but on balance I think the above holds.

It is also possible to appropriate an established community by bringing in new members over time with views opposing the founding principles. This is much easier if the leadership preaches tolerance.

This is one of those things that is kind of hard to say without people getting triggered because of negative stereotypes but sometimes you have to stand up for principles and kick people out of social groups to keep a good thing going.

I’ll never cease being amazed at how much people whine about scrolling past a post they don’t want to click on. It’s the tech equivalent of “my kid shouldn’t have to see ‘those’ types of books on the shelf in a library”. At the time of writing this, 16 of the top 20 stories on my feed are not AI related in any way.

I don't think the poster has the power to split HN in twain.

I don't think the poster believes some kind of democracy could bring about this.

I do believe that by entertaining the idea, the subsequent discussion will be useful for moderators to get a feel of what their userbase thinks of the current state of things.

From my understanding, the soul of HN and what makes it what it is is the moderation - having discussions on issues is an efficient way to signal to them.

> To me this seems like the subtext is "Those people are not welcome here, they are not like us. It's not like we have anything against them, we just don't want them ramming it down our throats"

I am truly tired of AI being rammed down my throat, not just via the tech news, but in article content (slop), in un-asked-for tech product features, and at my own tech job. The solution is not to divide the community and make people unwelcome, but to provide at least some minimal set of filters and ways to opt out of the hype frenzy. I don't want people to feel unwelcome, but I do wish there was a way to turn the AI firehose off.

It must have been somewhat the same when chess engines started to beat human players. The chess community should be fairly divided about the usefulness of such a tool. After a while things settled down, and all players use the tool in some way or another.

Some chess players benefited more from the tool than others. I always analyze my games carefully with an engine after the game. After less than 10 years I managed to get from zero to almost master level. I attribute that to extensive engine analysis I put on my games afterwards.

The user needs to know how to use the engine, LLM or chess engine, when it makes sense to use it, what are the shortcomings of the tool and so on.

LLMs are game changers, and AI's ability to distinguish the signal from noise is marvelous. Will it be a game changer like it is now for chess, a very narrow game compared to everything else, remains to be seen.

This honestly reminds me of the crypto days from 2017-2021ish.

Literally 80% of the posts were about crypto and how we were going to experience some ground shattering revolution. There were so many posts about how all the topics are about crypto and how it is annoying.

Ultimately, all that noise and the billions of dollars poured into that turned into a meme if we're honest. Most people just buy/flip crypto or hold BTC that they'll sell when they double it after a year.

AI in LLM form is at least useful in many ways and in front of millions of people without any rugpulls and other shit, but due to their inherent limitations (doesn't matter how much python it writes and executes, half the time or more its wrong for any actual/meaningful work) I think the hype will settle in the next couple of years.

For myself, i often want to be able to just "shift views" on an existing community, rather than wholesale move to somewhere else that fits better.

I find I can do that with granular enough subreddits, or the (maybe old) feature in Twitter where you could group people you follow into lists and see multiple "homepages".

This for me has solved the issue of dividing community, which at the least from a practical level can be tricky.

Ive been exploring how to achieve this effect "on top" of HN lately, rather than by controlling followers, by popping a very simple AI filter on top that re-ranks it for me, and found it quite satisfying, but not sure what the ultimate value/usecase might be.

I would like the Rust evangs to have their own group. Thankfully AI came and saved us all from their terror

But I don’t see anyone saying group B isn’t welcome in both groups. It’s not as though you can only be in a finite amount of them.

At the same time, I never saw HN do this with any other trend so why with AI?

I highly doubt this is going to happen anyway.

HN has many VCs and startups and HN itself is backed by a VC firm, so I highly doubt AI news is going anywhere as it benefits many YC Startups currently levering this hype as well as VCs and other investors shoving cash into this.

Additionally, that there are also ton of vibe coders, OpenAI / Meta / Google folks here interested in this topic.

I'm afraid the only solution is for people to ride the cycle until it either fizzles out or morphs into something else.

> Anyone is free to make a website with whatever content they want, they can invite people to it and grow your own community.

This is very hard to do. But hey, I'll give it a try.

Starting now a new community for AI-assisted coding: https://kraa.io/vibecoding

It's about a topic not the people.

I updated it to fetch 200 stories instead of 30, so even after filtering you still get hopefully 140+ things to read.

https://github.com/simonw/tools/commit/ccde4586a1d95ce9f5615...

That exclusion filter seems to be just a very dumb substring test? Try filtering out "a" and almost everything disappears. That means filtering out "ai" filters out "I used my brain not a computer".

Top story: Kiro: new agentic IDE

An interesting example of both LLMs' strengths and weaknesses. It is strong because you wrote a useful tool in a few minutes. It is weak because this tool is strongly coupled to the problem: filtering HN. It's an example of the more general problem of people wanting to control what they see. This has existed at least since the classic usenet "killfiles", but is an area that, I believe, has been ripe for a comprehensive local solution for some time.

OTOH, narrow solutions validate the broader solution, especially if there are a lot of them. Although in that case you invite a ton of "momentum" issues with ingrained user bases (and heated advocacy), hopelessly incompatible data models and/or UX models, and so on. It's an interesting world (in the Chinese curse sense) where such tools can be trivially created. It's not clear to me that fitness selection will work to clean up the landscape once it's made.

Probably would work better as a userscript, so you don't have to rely on a random personal website never going down just to use HN. I don't have a ChatGPT account but I am curious as to if it could do that automatically too.

There's a special kind of irony to use AI to help out the people who hate AI.

It's not hypocrisy or anything negative like that, but I do find it amusing for some reason.

This is neat, but with the given filters you autoselected (just the phrases "llm" and "ai"), of the 14 stories I see when I visit the page, 4 of them (more than 25%!) are still stories about AI. (At least one of them can't be identified by this kind of filtering because it doesn't use any AI-related words in its headline, arguably maybe two.)

This is interesting, but I found it amusing that you used:

"I built..." and "o3 knocked it out in a couple minutes...", not ironically, talking about a tool to keep us from having to be inundated with AI/LLM stuff.

I also built https://lessnews.dev (HN filtered by webdev links)

One decision I had to make was whether the site should update in real time or be curated only. Eventually, I chose the latter because my personal goal is not to read every new link, but to read a few and understand them well.

Almost certain you can use the HN Algolia to do the same thing by excluding terms

feature request for OP: sort by "LLM Agentic AI" embedding cosine distance desc

AI solving the too-much-AI complaint is heart-warming. We're at the point where we will start demanding organic and free-range software, not this sweatshop LLM one-shot vibery.

Love it. :D

It only shows 13 stories? And no pagination.

simon how do you get so much done? It’s incredible. Would love to see the day in the life TikTok :P

I think there is a fundamental disconnect in this response. What the user is asking for is for a procedural and cultural change. What you’ve come up with is a technical solution that kind of mimics a cultural change.

I don’t think it’s wrong, but I also don’t think we can really “AI generate” our way into better communities.

You can even make it live with SSE/EventSource.

[flagged]

Great example of the power of vibe coding. The first item is literally "Kiro: A new agentic IDE".

[flagged]

[flagged]

[flagged]

Perhaps you should add a privacy policy or just release the source rather than assume people will trust your site. Why do you do these demos if you aren't upfront about all the things the LLMs didn't do?

> Hanging out on the "new" page and upvoting quality non-AI articles is an effective method of resistance.

Fully agree, and I in fact am finding that I actually find more stories I'm interested in that way than looking at the front page. For whatever reason, I'm increasingly getting out of sync (interests-wise) with broader HN. So many stories I think are great HN material (and would have been a few years ago) languish with almost no activity.

So there are two reasons IMHO to browse new: Surface better stories to front page for engagement, and find better stories

>This post is turning up at least every other day.

Res ipsa loquitur

>Hanging out on the "new" page and upvoting quality non-AI articles is an effective method of resistance.

I hang out in /ask and /asknew for my part.

PS: Hey, Paul... When are you going to close my 2021 issue[0], you already merged the pull request[1] :D

Come on, man!

[0]: https://github.com/paulhoule/gastrodon/issues/10

[1]: https://github.com/paulhoule/gastrodon/pull/11

> The last few times my reply was "AI is 4/30 or 5/30 of the front page, it's not such a big deal", but today it is 9/30.

A bigger impact for me has been the number of mentions of AI in the comments. It's not just that a large part of the front page is dominated by LLM hype posts, it's that every single post has a least one guy near the top somehow bringing AI into the discussion. I don't even care if it's "AI will fix this" or "haha, AI sucks at this too". I just don't want to hear anything about AI ever again.

[flagged]

That was very different. Somehow the entire front page was Erlang, but it was only for a day or 2. AI is different from that. It's like a good 40-50% of the posts for at least a year or more, and I don't see it going away anytime soon. It's also different from web3/etc. as those were at most 10% of the posts and most of us can see it's just hype.

I'm not fighting for a split/fork, just stating the fact that it's nothing compared to Erlang.

It all depends if you care about the tech side of HN or the startup side of HN. I love the tech articles above all else and could easily do without the general trend fluff.

With that said, I don’t find the AI posts nearly as bad as the Blockchain era.

I wouldn't mind the Erlang-dominated front page coming back :)

Erlang is kind of a special case, since there was that period when the community's preferred response to "too much politics" was to spam submissions about Erlang. Agreed though, it doesn't seem to have taken over more than (say) Bitcoin or Rust have at times.

I miss the days of daily Haskell posts.

I've been here a while and this is certainly more, prolonged, and has no end in sight compared to most other hype cycles we've experienced.

It's also exceedingly generic such that AI isn't really a topic, it's an entire classification or maybe domain to steal from the animal kingdom hierarchy.

Microservices had a micro moment not much longer than xml.

The best one was the 2048 era: https://hn.algolia.com/?q=2048

I hope I'm not the only one here who never heard of Erlang until I read your comment (I arrived in 2016).

Oh my god, the years of the JS frameworks. Millions traumatized for life

One might argue that those technologies are sunsetting now

Yes, but unlike AI & Crypto, Erlang came with little grift, slop and Show HN spam.

Can you link to some specific examples of low-quality discussions?

As with any Major Ongoing Topic on HN (https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...), the goal is to reserve frontpage space for the higher-quality stories and try to downweight the follow-ups and low-quality ones. We can't do this perfectly, of course, but we try.

I feel it's to generic in application to be interesting to a broad audience like HN. Some things I like because I have interest in the problem space and am interested in how they applied AI to it. But most things I'm not even interested in the problem space and so could care less how they applied AI to it.

[flagged]

Wouldn't be hard to train an LLM to do it!

Someone has probably developed a personalized browser for HN.

I don't know if you're allowed to promote alternatives to HN, but Lobste.rs has tags which you can follow or completely block. Plus I've found the quality of the discussion is higher than HN, at the cost of much lower quantity.

That is a perennial proposal as far as I remember.

I second that emotion.

Keyword filters on the user side would avoid adding extra steps to submitting and moderation. Categories are an extra complication.

The problem is the quality, not the topic. Understanding serious papers about AI development requires fairly specialist knowledge; there are plenty of people around (like myself) who have been programming for decades and can write really nice code in a bunch of different programming languages, but have very little if any mental model of "transformers" or whatever.

So in practice, "AI" content ends up revolving around people bandying about opinions about whether or not we're all doomed, or whether or not we're all on the edge of a utopia, or how much productivity programmers (and which ones) have lost or gained, or what kinds of tasks the LLMs are or are not currently or still good at, or whether anyone still cares about the fact that the term "AI" is supposed to mean something broader than LLMs + tool use.

The emergence of the "vibe coding" concept has made things worse because people will just share their blog posts about personal experiences with trying to write code that way, or flood the Show HN section with things that are basically just "I personally found this specific thing to be 'the boring stuff' that's actually relevant to me, so now I'm automating it" with a few dozen lines of AI-generated code that perhaps invokes some API to ask another AI to do something useful.

Why kick it out, in the past when similar annoyances of dominating the front page occurred they created the Show link and the Ask link. For people interested in those they still exist, just away from the front page

A lot of people on the internet have turned hating AI into a personality.

It means there are grumpy curmudgeons in every community.

[dead]

> There have been some reddit communities over the years doing this, but IME they frequently devolve into echo chambers. It almost always comes down to bad moderators.

In my experience:

Most of them are basically designed to be echo chambers from the start — opposition is only admitted in to the extent that it allows easy targets to knock down. Most people just aren't that good at explaining why they believe what they believe, let along making a convincing argument for it; so all you need to do is set up an environment where one side's position is the default.

There have been a few attempts at explicitly avoiding that problem. They do eventually collapse. But I don't think it's due to bad moderation. It's more that certain factions simply refuse to engage civilly and unemotionally with each other. They will see statements as inherently provocative that the other side genuinely consider matter-of-fact.

I was a moderator for a place like that once. It was remarkable to me how, on the "hot topics" that were polarizing and led to a lot of bans and suspensions, on one side people who were suspended would argue and whine and complain basically as long as we'd listen to them, maybe even the entire duration of the suspension, and they would never get it into their head what our standards were for respectful discourse; and they would even suggest that having such standards was inherently oppressive; and when they got back they would immediately go back to their old ways. And on the other side, people would basically say "LOL, see you on <suspension end date>" and disappear, and come back as promised, and behave themselves for a while.

And while there were a very few people who simply couldn't kick the habit of using slurs or other disparaging terms to refer to identifiable groups of people, there were far more — almost all on the opposite side — who simply couldn't kick the habit of openly insulting the people they were directly responding to. Or at insinuating negative character traits and hidden motivations not in evidence, or other such "dark hinting" as we call it. Or even just of using obnoxious, brutal sarcasm all the time when we expected people to speak plainly.

I have a similar process, but usually scan down to 60 (as I did today). I found eight stories including this one and have tabbed them to read. I don't like rust-y koolaid myself, but would never complain that it is here, nor would I complain about seeing the word 'typescript' as it is really far from my interests. To my interest -- excellent AI related white papers, AI agent paradigms and code, model announcements etc. are regularly posted here. Of the eight I picked today, half are AI related. 4 out of 60 isn't bad if I was trying to be an artificial intelligence ostrich.

Very few problems were solved by crypto, so it naturally disappeared.

On the contrary, LLMs based AIs create a lot of new problems.

Unlike crypto though LLMs are actually useful.

But but the Rust stuff is still there ! (To my pleasure I must confess)

I haven’t noticed any slowing, if anything it’s accelerating as people try vibe coding and realize they can build an mvp and get some suckers to sponsor on GitHub. Just look how many end up with donate links. I suspect a large portion of people releasing open source vibe coded projects don’t care about the project but see it as a low effort way to make a few bucks.

> Two, fragmenting the community would just reduce engagement and risk making both feel like a ghost town.

I would rather hang out on a HN more similar in attitude and user count as when I joined than have it be "lively" with the user base it has now.

I would agree with #3 if it were the case that the technical side of LLMs were discussed more, and there was a focus on trying to run and understand them yourself; these are topics that I think make more sense for hackers to care about. The front page stories about LLMs are usually business news about LLMs, blog post about using (someone else's) LLM, the occasional blog post that says LLMs suck either for technical, social, or other purposes.

Obviously I'm missing a few categories here, but my point is that there is basically zero hacker spirit in the LLM posts. Since when is it part of the hacker ethos to plug into a SV machine that you pay $200/month for the privilege to use to do stuff? That's ridiculous...

It's hard to not feel like this is part of a bigger transformation of the culture and then absorbing that culture on HN. Hackers make things, and they are unlikely to settle for poorly made things, but routinely the chorus on HN is that you should never reinvent the wheel, etc., even though you could make a better wheel for your purposes than the wheel you've found online. Not to mention this hilarious "I hate writing code" sentiment that seems to have grown over time. In my opinion neither of these sentiments belong in any group claiming to be hackers.

But the user base of HN is not hackers, hence why they feel comfortable spouting these opinions, plugging into $200/month services that write code for them (badly and slowly, I've used Anthropic's Max plan with Claude Code), not understanding and/or taking apart the technology they use, etc.

Difficulty - AI makes it a lot easier to generate mountains of hype articles and drown out other content, self-reinforcing the hype.

Different than prior hype cycles.

Frankly, this one seems to be dying out more from everyone just flat out refusing to pay attention to online stuff or things on their phone long enough to starve the beast. If that is even possible.

But finding out that LLMs are used to push the political posts in such obvious ways yet people are still falling for it is a somewhat interesting topic. Regardless whether that just pushes your bias towards LLMs == bad or tweaks you to spin up your own LLM to combat by searching/detecting/posting the slop. Burying your head in the sand and ignoring all of it is not good either.

The desire is relateively simply explained. Some people used to find HN interesting, but the modern set of things being upvoted isn't matching their interests anymore. They already ignore the content they don't like, the problem is the content they do like isn't there anymore. The assumption, I expect, is that if there was less LLM content the site would have more of the "older style" content they used to enjoy. I don't necessarily think that will happen, but that's my interpretation of the sentiment.

It seems that, for this to be true, AI (not just LLM) will have to plateau, very soon. Considering the Nobel prize for chemistry, last year, was about AI (not LLM) [1], I think there are still an incredible number of non-LLM applications of AI out there still.

[1] https://www.nobelprize.org/prizes/chemistry/2024/press-relea...

I feel like 10 answers is too few to generate a good profile. I tried it and it marked a lot of stories as 'skip' that were pretty interesting to me, presumably because they didn't fit into the 3 or 4 categories it determined I was interested in. That said, it was useful to filter out AI stuff, so that's good I guess.

The mere presence does seem to a offend. HN has what seems to be on average a more negative appraisal of AI than the industry average (based on where new software jobs/funding are going), but some of the comments I read seem to imply that this page is swamped by endless AI hype posts devoid of substance.

My first instinct reading these comments is "are we on the same website?", but I realize their perspective is possibly skewed by a strong visceral dislike of AI as an affront to many fundamental things they like about software and tech. It's become a tribal conflict, based on the ethos of "whose side are you on" rather than a sober appraisal of the facts, benefits, and legitimate hazards.

When a quarter or more of the front page is stories about those themes, you'll be doing a lot of whacking.

Thank you for this - works like a charmandno need for a github project like the one proposed above.

Or just spend 60 seconds writing a ublock filter with the ten most common phrases you don't want>

Perhaps you can fork and enhance https://github.com/panicsteve/cloud-to-butt

From time to time, you find something AI/LLM related that is interesting. Like the people trying to save money by speeding up their audio before submitting for transcription. Not only did it save money, but improved accuracy. It's those kinds of finds that are interesting in the hacker sense. Increasing the number of tokens at a certain point has a negative affect. Okay, someone is taking that hacker ethos to see what happens when the knobs are twisted. Showing me yet another image manipulation tool or new chat bot? Yawn. Next.

I'm seeing quite a few plugs for lobste.rs ITT. My recollection of my impression from many years ago is that they had different politics from HN (I did read it every now and then, even if I only actually joined months ago) of the time, but that many posters were still very much engaged in ideological battle, and furthermore flagging and downvoting was used more to suppress one side of the argument than it was to keep the site tech-focused.

You don't even have to create it. https://lobste.rs exists and is opensource https://github.com/lobsters/lobsters It even has tags built-in.

Its pretty easy to do the user-loyal bit, with a bit of prompting to give an llm your preferences/profile. Not ideologically loyal, but i mean acting in accordance with your interests.

The tricky part is having that act across all sites in a light and seamless way. Ive been working on a HN reskin, and it only is fast/transparent/cheap enough because HN has an api (no scraping needed), and the titles are descriptive enough that you can filter based on them, as simonws demo does. But its still HN specific.

I dont know if llms are fast enough at the moment to do this on the fly for arbitrary sites, but steps in that direction are interesting!

Now that is a browser I'd pay for. A genuine user agent, rather than just a browser.

It would also need to be able to "Recognize tasteless, ad-ridden, or other difficult-to-read pages, silently dismiss cookie popups and signup solicitations, undo any attempts to reinvent scrolling, remove all ads except for those on topics X, Y, and Z, and present the page using something like Firefox's reader mode."

Other requirements would include "Fill in these fields that are marked as autocomplete=off," "Use this financial site to display exactly the charts and tables that I want, in this order," "Clean up broken, irrelevant and repetitive search listings on Amazon and eBay," and so on.

For extra credit: "Maintain this persona on Facebook, this one on Bluesky, this one on Slashdot, and this one on HN. Synthesize documents needed to establish proof of age and other aspects of personal identity."

> as long as the AI/LLM posts are not dumbed down mainstream content

The problem is, most of it really is (as it boils down to "I am / $COMPANY is using an LLM to do something; here's how you can do it too, and / or some pundit's opinion of the implications for the industry"). And the stuff that wouldn't be (like how they work, or statistics and benchmarks), often requires relatively specific domain knowledge to really appreciate.