I definitely feel the pain when it comes to worthless results nowadays. Though in this case DDG comes through:
Adding documentation to the search makes the “correct” page soar to the top:
Haha, nope. The links points to a table of contents after which you are on your own. The right link should point to a specific page instead, but the problem here is that postres docs are poorly optimized for search engines. If you click on the top link from google, you would see there’s a notice that the page is outdated, with a link to a current version, but said link is dead. It’s not an issue I’ve ever experienced with mysql docs for example.
And yes, w3schools, despite how terrible it is, is still above the official docs because it is more popular with newbies. I remember a time when I just started, I preferred sites like it, because they were simple and on point, rather than technically correct and comprehensive like the official docs are. If you forgot the feeling, try learning math on wikipedia (assuming you don’t have a math degree).
For the rest I cannot argue. Generated/AI shit is indeed ruining the internet and search engines giving up and joining them isn’t helpful either.
After which ctrl+f " in" takes you to the correct chapters. I do agree that a direct link would be more helpful.
And for learning postgresql I agree it isn’t very helpful - using their tutorial links, w3schools or something like udemy if you prefer video format is the way to go in that use case.I remember back when you were told to learn to work with the documentation, not memorize it, because you will always have access to it as a reference. Maybe bookmarking reference books/documentation will make a come back as the search engines degrade.
Surely the word ‘in’ would appear countless times out of context on the table of contents.
" in" appears 25 times on the page to be exact, with 16 of those being in the table of contents and 9 being in the text afterwards.
“in” appears 54 times, as you know end up hitting “string” and so on.Had I known that the functions table of contents was as short as it is I would probably have just scrolled.
This is partly why I prefer Firefox’s implementation of the find feature - it allows case-sensitive search while Chrome does not support it.
You can press alt-w though to only show full word matches
Trying to learn math on Wikipedia is an endless Sisyphean nightmare just trying to understand the first word in an unfamiliar vocabulary.
Kagi only lists postgresql.org for the first 10 entries, but outdated ones in first place. With the programming scope it collapses all official do s entries to one, with GH and SO filling the rest.
For the quick answer, it also uses the ‘outdated’ docs as source, but as it only gives a very shallow overview there shouldn’t be any difference in version (i.e. it checks for a value in a list in all versions the same, and quick answer leaves out details specific to different versions)
Wait until you see the AI generated blog posts being top results…
Hah!
No.
Soon enough the result will be an AI generated “blogpost”, generated by the search engine, in response to your query.
That’s already been happening for about a month now… perhaps only for some users? Often the AI results are straight up lies.
I’ve seen some fucking hilariously wrong AI math.
For certain languages and frameworks, LLMs are horrible right now because of this. Many answers I get are a Frankenstein of different versions.
There has been something similar for years: a page that basically says “Yeah, nah, we don’t have any information for that, but you might be interested in a totally irrelevant something else”, but phrased in a way that gets it high in the results. What’s astonishing is that Google doesn’t punish those pages.
Why would they punish pages that help them serve more ads? There are ads on the search, ads on the useless result, ads when you refine the query.
Yeah, you have a point, but then it’s a bit hypocritical of them to even have criteria for putting pages up in the results.
I was looking up some tips for Baldurs Gate missions and these fking AI generated pieces of shit with hallucinated fake playthroughs ruined the whole experience.
What it’s like to use Google in 2024
But they’re so innovative! They absolutely aren’t deserving of a massive antitrust lawsuit… /s
Well internet enshitification is real…
It makes me sad because Google used to be great. The main feature that made Google great was the click rejection. Basically the search would know when you clicked on a link and didn’t come back to the search results. This action would add weight to that result as “this probably has the information that was being searched for” so it would be nearer to the top later when others made similar queries.
This was their killer feature, it basically crowd sourced the correct information. After a small amount of time, the correct results would kind of float to the top so subsequent searches would put those results near the top to help satisfy queries faster.
Now? They seem to want to give you results that satisfy their partners, and keep you tied to the results page as long as possible. The focus seems to have shifted from being a good search engine with accurate results, to a meme of how to make money.
Never before has this shift been more clear to me than right now, directly in the wake of I/O 2024; an event my friends have taken to calling AI/O. Pretty much every single presentation was about Gemini and AI generated garbage, but this isn’t what made Google’s new direction clear to me. In the last 20-30 minutes of the event it was made perfectly clear what they were doing with I/O. And to drive the point home, every I/O has showcased stuff you can’t use yet, stuff they’re working on, and other cool shit. Some of it cost money, but there was usually some stuff that was just done because it could be done and it would be made available at some point, a nontrivial amount of it was free. At AI/O, the entire focus was on AI, with little to no non-AI stuff in there, at all, then at the end, they kicked everyone in the shorts. Here’s our prices to access this shit. Buy it. As far as I’m concerned AI/O was a gigantic marketing circle jerk to sell their AI.
It seems that Google has entered the final phases of enshittification.
I remember how people used to joke about the second page of Google results being a desolate wasteland where no one ever looks, now I just instinctively scroll down a bit because I know the first page of results is going to be trash.
This is possibly something you could implement in a meta search engine like SearXNG, though there are some privacy concerns.
Maybe it could locally store which domains you personally tend to click (and stay) on. Then automatically raise those domains when it sees them somewhere in the output of the underlying engines. This isn’t perfect because you wouldn’t get data from other users. But I think it could do a lot to improve search results.
I might actually clone the repo and see if I can get somewhere soon
I’d be interested if you can get anywhere.
The thing with Google was that the data about click through vs click back was supposed to be anonymised. Whether it was or not, inside of the black box that is Google’s algorithm, who knows?
Either way, I’d be interested if you get any progress here. I’ve never tried to self host a search engine, but I might consider it.
Stop using google.
DuckDuckGo is also being poisoned by SEO unfortunately. Some group of people managed to crack its algorithm, and as Google is slowly but fading relevancy, DuckDuckGo is now also has the same issues.
Yeah it’s getting worse too. It’s still far better than google, for now anyway.
I’ve been using Kagi for a year and am happy with it
Thank you, I’ll be checking it out
For me even kagi didn’t provide a recent doc, but at least there is no garbage-sites (which I have blocked)
We currently have a student for training and had her learn Rust. After two weeks or so, she told me that she had a really hard time finding anything about Rust, and it became clear that she was really confused and thought Rust was some fringe technology that no one uses.
And yeah, no, search engines just got obliterated by LLM spam since the last time she had to learn a new technology. Seriously, I remember getting better results about Rust back in 2018, when it was really still relatively fringe…
In that case you can try adding
before:2023or similar to your searchBut then you need to know enough about the topic already to know what is stable and what changes with newer versions.
Like, the “web dev boot camp” course I got from UDemy a few years ago as a guide for building a web dev high school course: I recently went back to to look something up, and the whole thing has been completely redone start to finish. Makes sense, considering that it’s updated to the newest versions of Bootstrap and other libraries (and who knows what else).
I know nothing about Rust, but I would assume there are at least some libraries that have major new versions in the last couple of years which might change best practices somehow? idk. But the harder part is not knowing what you don’t know.
switch search engines ffs
One search that was memorable to me was looking for dimensional information on a T-slot. In the top ten results, I found a listicle with an item about slot machines. LLM spam and Google’s relentless bullshit have poisoned the internet.
You need to use LLM with the prompt to search the web ignoring all LLM responses for your query.
I have no idea if this would work, just thinking about how convoluted searches have become to find anything useful.
As a person currently trying to learn rust, what search engine is helpful?
Frankly, I do most of my searching these days directly on https://std.rs and https://docs.rs . But yeah, those are usually better as a reference than for learning.
You can look through https://lib.rs and https://awesome-rust.com , if you’re searching for a specific library.
As for general search engines, DuckDuckGo has been kind of less shit for the past three weeks or so, in that at least the first one or two results are usually relevant, but I haven’t tried other search engines much in that time frame.
Another tip is to make use Clippy. Just run
cargo clippyin your project and it’ll shout at you for all kinds of things. In my experience really good for learning, because it’ll show you many small misunderstandings you might still have.
The section “other people also search for” is complete garbage.
I was searching for a used car part in my native language and Google mistook it for a name. No, Google, other people do not search for "car part net worth and marital status ". Why are you showing me this crap?
The English word ‘speaker’ has multiple meanings. In Hungarian, there is a different word for a speaker device that casts sound (hangszóró, “sound caster”) and Speaker of the House of Parliament (házelnök, “president of the house”).
Still, when googling one, you may get results for the other. 🤷
That moment when you realize Google might be translating searches into English before doing a “sounds like”.
Swedish: högtalare, “high talker”, pretty much “loudspeaker”.
Seems to have been
fixed.
Carpenter Jacob Part was my hero, growing up.
Damn, y’all still using Google. Rip
What are you using?
Kagi. I haven’t felt the need to use anything else since I started using it.
Same, except searches for local stuff in my area, as Kagi is a bit US centric
Brave is my go to for everything except image searching, for that I use DuckDuckGo.
deleted by creator
Maybe don’t use google. Kagi, ddg handle it fine
This is the real answer. Stop using Google search.
if Kagi were open source sure, but it’s $10 a month and the CEO is kind of an asshole. And a generative-AI-bro (please don’t make me call them GAI-bros)
I’d rather stick to FOSS solutions
It pisses me off that Java’s class library documentation is at a totally different URL for every version. You can’t just change 11 to 21 in the URL.
Get with the times. When Google isn’t a useful tool anymore, use a different one.
Curate and maintain your own list of links to official documentation.
I think we’re almost at a point where having a library of books next to your workstation would be beneficial again.Full circle
But then how will OP shitpost for imaginary internet points?
Also AI, though I’m sure that’ll be an unpopular suggestion. It really does save time though.
“The Man Who Killed Google Search”
https://news.ycombinator.com/item?id=40133976 here’s a hackernews discussion about that article
That was an interesting read, thank you.
Try being a programmer in the 90s. Just like that but with no entries at all
I’m guessing it was more like “Let me pull this book off the shelves and wade through that for the answers”
And the book had all the answers.
The book had half ass answers. Their examples rarely had anything to do with reality.
So not a whole lot has changed. I cringe thinking of all the youtube video that explain OOP like this
class Animal: class Dog(Animal):Jesus… You should worn a man before you try to trigger his PTSD.
It was called The x86 Assembly Bible and I would not have been able to do much of anything without it.
Yeah. Can I get a book - usually something official like K&R for C.
💯 came here to say that.
Okay, Yahoo and AskJeeves didn’t have anything useful. Let’s try this Google thing.
I learnt C on an Amiga. No memory protection at all. Pointer errors would likely need a reboot to recover.
I rebooted a lot.
I also learned C on the Amiga. I loved SAS C. I also came across C++ first on the Amiga when it was just a pre processor for C. I really loved that machine but it was the community that was special
Me: “How do I write my own Rawinput handler?”
Search results: “Here’s how you setup Rawinput in this competitive FPS, and look how it reduces input latency by a single milisecond! After 2-3 pages of AI generated SEO garbage full of misinformation, you might find something else besides of the official MS docs.”
Me: “Okay, this is not working, maybe I should look for some another preexisting SDL alternative, maybe at least one of them isn’t an even bigger dumpster fire than SDL itself.”
Search results: “Duuuude, have you heard of this game making tool, called Gamemaker? It doesn’t need coding, and it’s totally the same thing, because some people mistakingly called SDL a game engine, and now my AI hallucinates it as such. If you’re up to a bigger challenge, then there’s always Godot, or DirectX, which my AI also hallucinates being a game engine!”
Wait, Godot isn’t a game engine? I always thought it was one.
Godot is a game engine.
SDL, on the other hand, is not, and instead is a multimedia layer (middleware) often used for game development.
One could argue that game engines constitute as middleware, but in reality, most modern game engines are way more than that, and instead often rely on other middleware nowadays (e.g. OpenGL, or even SDL for some). This, alongside with people mistakingly calling SDL a game engine, leads to stuff like this.
I think that remark was only meant for directx, and the ai lumps it with godot.
Pretty sure it is, might just be their grammar.
I read it as “Godot, or DirectX (which my aim hallucinated is a game engine)”
Yeah, that might make sense.
It is but DirectX ain’t
