Data on AI-related Show HN posts
ryanfarley.co
ryanfarley.co
(Caveat; not sure this kind of flagging is a pattern)
Link to the original article: https://www.the-independent.com/tech/chatgpt-psychosis-ai-th...
Related to the article, are there actively updated benchmarks for plan recognition?
Only a subset of people were, and still are were involved in Crypto, whereas LLM is of use to pretty much everyone.
As a seismic change, of course it's going to be everywhere till the hype cycle flattens.
"show hn" "nft" - 151 results
"show hn" "blockchain" - 479 results
"show hn" "crypto" - 782 results
"show hn" "llm" - 2,363 results
"show hn" "ai" - 13,128 results
trends or trendy does not imply substance.
Most of the time, AI tools promise to be timesavers. So it's natural many folks look for shortcuts. We're simply overloaded, partly due to current situations generated by existing machine learning tools deployed elsewhere in the system.
Your statement is also why I fear this supposed promise that "AI will do all the work, society won't need jobs!". I don't think we're getting this post-work utopia that tecunocrats love to promise.
I didn't use it much growing up since they moved west when I was young. but it turns out that "y'all" is surprisingly nifty: a gender neutral, 2nd person pronoun for a group of peope. So I picked it up more in adulthood and put it into my daily vernacular.
And also by Indian Christians (Catholics) in some parts of India, such as Mumbai and nearby areas, like Pune and Goa, along or near the Western coast of India. Partly grew up there, and also did some of my schooling there, that's how I know this.
I don't know if there is any historical connection between the usage of that phrase (y'all) in those two areas (of the US and India).
It could have been, via (US) Christian missionaries coming here. There were and still are some of them, in some parts of India, from more than 100 years ago. Again, I know this from experience.
That general area of India, and some other parts, do have a relatively high percentage of Christians.
Yeah(, right). Or rather, yeesh. Or maybe, yikes!
>we all just have way too much to do. Every day. All day.
... in the richest country in the world, which some inhabitants there also call the greatest nation in the world.
Being a blog it also has RSS which you can use to consume it in the reader of your choice.
They feature projects done by people from across the web so they refer to people’s names using square brackets e.g. [josvdwest]. Each post has comments which I find to be quite thoughtful.
There is also https://hackaday.io/ which is a repository of hardware projects. In very simplified terms it’s like a cross between Facebook and GitHub for your side project that involves custom hardware. People post the details and their progress. It ranges from half baked attempts that were abandoned early on to some really amazing and unique things. When people complain that the old web is dead, this is the antidote: just clever people creating and collaborating on cool shit.
Maybe you are fine with corporations doing that, but imagine now if instead of corporations they were an alien race doing all that to us, would you not be worried for our future and autonomy?
Quantum computing is progressing slowly but it's most likely going to be mainstream, yet too technical for the average person to care about it.
Today’s tools like Spanner are vastly more sophisticated, but were built by people who learned to work at petabyte scale developing in Mapreduce
We’re getting better AI tools every month, and the best way to be ready for next year’s tools is to work with the tools we have now
a) Most date/time functions work with TIMESTAMP, no need to convert to/from DATETIME.
b) You can do case-invariant regexes for REGEXP_CONTAINS() by prefixing it with (?i), which is more performant than casing w/ UPPER/LOWER beforehand: https://stackoverflow.com/questions/42987537/google-bigquery...
or a more complicated prompt to include "machine learning" or computer vision, etc
or another method... such as filtering on tags is there a HN viewer with tags...?
found this HN post from 2 years ago https://news.ycombinator.com/item?id=35904988
bad-ai.js
const bad_words = [ " ai", "ai ", "a.i.", "openai", "llm", "agentic", "claude", "large language models", "apple intelligence" ]
var elements = document.body.getElementsByClassName("titleline");
for (var i = 0; i < elements.length; i++){ var element = elements[i]; var text = element.children[0].text.toLowerCase();
bad_words.forEach(word => {
if( text.indexOf(word) > -1 ){
console.log("Bad-AI remove: " + text)
element.parentElement.parentElement.nextSibling.style.display = "none";
element.parentElement.parentElement.style.display = "none";
}
});manifest.json
{ "manifest_version": 2, "name": "No-Hacker-AI", "version": "1.0",
"description": "Hide stories on HackerNews, if they mention AI technologies.",
"icons": {
"96": "icons/bad-ai.png"
},
"content_scripts": [
{
"matches": ["*://news.ycombinator.com/*"],
"js": ["bad-ai.js"]
}
]
}Meanwhile, I really liked the blog theme/design. (reading on an iPad) Simple, yet powerful, with a nice touch of the blinking underline at the end of the title. (It's a simple trick via CSS, but nice to see!)
Also, 1 in 5 being talked about is different from what is actually getting funding.
I’d say you can only say saturated relative to the historical popularity of other topics on HN throughout the years. There’s no real golden rule here
Like it or not AI is not only a huge trend, it really is a paradigm shift in all sorts of aspects
Would be curious to see the breakdown of this statistic
Well, you apparently just did. Saturation is indeed relative (or rather, on a spectrum with varying subjective lines). You can explain why it's saturated, but that still means it's saturated.
>Like it or not AI is not only a huge trend, it really is a paradigm shift in all sorts of aspects
Sure, so was fracking. I wasn't in this community back then, but I'd wager somehow we didn't have 1 in 5 articles about how much better society can drill oil.
I was literally expressing curiosity and wondering what the frequency relative to past topics and trends is and totally open to it being actually saturated
I didn’t define or explain what saturated is, I said it’d be good to see those stats - you don’t know the answer as you’re obviously conjecturing about fracking popularity quantified on HN but don’t actually know anything regarding those numbers (you might now after you go look them up to reply to this, nj)
And fwiw hn includes a lot of article types but it’s obviously already slanted towards tech so if you don’t see how much closer ai is related to tech and why it’d be more popular than fracking then I don’t know what to tell you
If anything, I should have noticed that they are talking about show HN because that does change things but I’d still be curious to see stats
Like it’s not that serious I don’t know why you got so offended
Either way, not interesting in engaging with some hn smart ass atm so have a good day and feel free to have the last word
Last thing I’ll say is HN is so fickle, have no clue why I’m getting downvoted
Sometimes I feel like it would be great if it were required to enter a reason
If you took my reply that way, I don't know what to say. I'll respect your caustic retort and cut off the conversation. HN isn't the place for flamewrs.
>Last thing I’ll say is HN is so fickle, have no clue why I’m getting downvoted
Note that I cannot downvote a direct reply. So I am not responsible for those actions.
I guess that’s because that’s where a lot of VC and hypebuxx are right now, though.
In 10 years either advanced AI will have eaten everything or nobody will even remember any of this.
My position is that I think AI is an amazing tool, but that’s it. It’s like any other tool we humans have developed before – we use it to create more quickly, with less effort.
Personally, I don’t think we’re anywhere close to achieving AGI, and I find the tech bros’ AGI fantasies appalling.
It could also explain the lower votes, "AI" or "GPT" being more generic terms is correlated in my personal experience with lower quality.
I think the methodology is too poor to actually make that call. It's closer to "the 20% is kind of made up".
Would it be possible for the moderators to penalize such articles so they can't easily gain traction similar to articles about politics?
When large emphasis is on the technology (whether it's made with AI or written in Rust or whatever), it seems at least in some cases like there is less substance to the idea itself, as a good product is a good product regardless of how it's built.
A post like "Show HN: I'm using AI to sort a list of integers" is inherently boring since if you remove the AI part it's something we've been doing for half a century.
A post like "Show HN: I'm using AI to translate Linear A" is interesting and would be interesting even if AI wasn't involved.
1. I love AI/ML hearing about stuff, and seeing it boom this much is great.
2. I really do enjoy working with LLMs and seeing what they can do.
3. It is quite amazing what non-technical people can now do with AI Assisted coding.
4. Working with LLMs within IDEs is getting quite good too.
I understand that there are still people who are not buying it, but quite honestly, its becoming harder to side with them. I have been in Software for 2 decades, and this "craze" has given me the most amount of enjoyment I have gotten since I figured out how to build a website sometime in my teens!
the AI craze is in full boom, feverishly stoked and fuelled by the hypesters and hipsters, who stand to gain from it.
reminds of the gold rushes of the last few centuries, like in California and the Klondike:
https://en.m.wikipedia.org/wiki/California_gold_rush
https://en.m.wikipedia.org/wiki/Klondike_Gold_Rush
picks and shovels, folks.
there's a sucker born every minute:
https://en.m.wikipedia.org/wiki/There%27s_a_sucker_born_ever...
I would have guessed higher.
AI is a seismic event for technology, venture investment, and eventually society.
I think this definitely remains to be seen. Might be, might not be. They were saying the same thing about the Segway.
Segway sold at most 10k units in a year.
ChatGPT currently has 77 million MAUs in the U.S.
>>Happens with all other specialized communities
I don't recall HN being specialised in AI, but maybe I missed something.
It's like posting your hardstyle remixes into /r/classicalmusic.
Okay, but what happened in 2018?
If you add "not made with AI" you can get counted in the author's totals for AI generated stuff
How do posts where the point of the thing is to "be AI" do compared to posts that just mention AI as a tool used?
To me, it feels like studying a new physical phenomenon. Like when Nicola Tesla was playing around with coils and wires, eventually loading to the creation of an entire industry.
Except, with LLMs, you don't need multi-million dollar equipment to play around with models. You can get pretty cool stuff done with a regular GPU, and even cooler if you use cloud.
I would say, if you are not spending some spare time fiddling around with LLMs trying to get them do some of the work you would otherwise do by hand, you are missing out.
Idk is 3th party actors or an push for "AI"
[1] https://bsky.app/profile/edzitron.com/post/3lsukqwhjvk26
Just asking for a friend.
Is there any estimate of how many comments are LLM-generated? What if I tell an agent "make an HN account and post comments with the goal of maximizing karma" and come back after a week to see how it went?
OK, I’ve done my bit.
In contrast, my random side projects that aren’t about AI get discussed here more than 50% of the time.
I don’t like it as well, but it’s a reflection of what people are concerning themselves with, so it is what is is.
i’ve been hiding a bunch of AI/LLM posts out of habit with the background idea that it might be useful for regexs.
My current filters are:
urls: twitter.com youtube.com
keywords: gemini openai claude llm llms ai agi
Make the front page more interresting. I'm on browser right now, otherwise this thread would have been filtered.
news.ycombinator.com###hnmain .submission:has-text(/\b(llms?|vibe[ -]?cod[ei]|claude|chatgpt|altman|agentic|ai)\b/i)
I removed most words to keep it short, it's easy to adjust according to one's tastes.
If before there was a bit of an attachment to checking what was going on on HN, now there's an overall meh. I still do visit out of habit, but when I do, I have this blasé attitude that quickly takes me away. Too much LLM this or that, AI everywhere.
HN being one of the last Social Media sites I've engaged with, it's good to finally let it go. Yay?
I get my "people doing interesting things" fixes on YouTube and Hackaday.
The problem with the current wave of AI is that it isn't "generally useful" yet. These AI systems can be very effective in certain situations when they are used with the specific knowledge of their flaws, but they are being applied to a much too broad use case today. Granted the valid use cases for crypto were miniscule in comparison, so I won't quibble too much with your general point.
Although the most frustrating aspect to me is that the people who were shown to be right about crypto have not earned any gravitas when they say similar things about AI. So many charlatans were saved by a second bubble immediately inflating as the previous bubble popped and we are once again ignoring the people calling them charlatans despite the proven track record for calling out this behavior. Seemingly no one learned anything.
Anything that gets posted to HN between when you’re fifteen and thirty-five is new and exciting and revolutionary and you can probably get a career in it.
Anything posted to HN after you're thirty-five is against the natural order of things.
I’ve been on HN since near the beginning (under a different user name originally). And there have been quite a few trends rise and fall on here. But none of them were as intense as this AI hype currently is.
In fact it’s not even AI in the more general sense, it’s almost entirely just LLMs that get discussed. There’s so much more going on outside of LLMs but all anyone is talking about on HN is natural language tools.
Generative image / video / audio models can produce output in image, video and audio. Those have far less applications than models that can output text, structured data and code.
"AGI is right around the corner" "No it's not" "Yes it is, LLMs are the future." "We don't even know if AGI is possible." "LLMs are the future." "No they aren't." "AGI is right around the corner..."
or
"LLMs are really useful." "No they're not" "Yes they are." "No they aren't." with a little bit of "They sucked the last time I used them." "Did you use them recently?" "That's what someone said last time." "But LLMs are really useful" ...
over and over and over.
It isn't even that it's mostly just LLMs being discussed, it's how they're begin discussed, they're effectively just a proxy for optimists and pessimists to argue over which worldview is better.
If we were talking about AI in general and not LLMs, the same conversation structures would still pop up.
When you look at other applications of AI, such as models trained for medical usage or AI used in Hollywood (something I personally have professional experience in) then it’s a completely different story because they’re bespoke models trained and used for specialised edge cases. Rather than Joe Blogs using an off the shelf package to do the same thing as the previous person albeit slightly differently.
I wish there was an HN of everything outside of tech.
(Of course, these sometimes overlap, but for articles talking about how tech/AI is helping solve medical problems, I’ll allow it.)