I like to read spicy things online. I’ve found that Google, even with safe search disabled, hides much of these unless your search is itself obviously explicit. I append “sex” to search for these without affecting results much. Entire sites are hidden otherwise.
but along with the change you noticed, there was another change. if you have safe search off, and add a spicy word into your search, you will be shown spicy things no matter what, many things that have nothing to do with the other words you used to narrow down your search because obviously you are interested in bewbz and god knows what else
The downside of this approach is that it can affect the search results returned. But I found that if you add " -fuck" or " -fucking" to your search term, it disables the AI summary without significantly affecting your search results (unless you happen to be looking for content of a certain kind).
You can probably find some other term that disables the AI but is unlikely to occur naturally in the articles you'd like to find, e.g.: "react swipeable image carousel -coprophilia".
Will it still work if "fuck" is part of a quoted phrase? If so, you could avoid it by constructing a phrase that contains the term but isn't going to match anything, ex: -"fuck 5823532165".
If I start fucking adding swear words to all my fucking search queries, how the fuck will the stupid ass search engine know that I did not want it to use that shit as one of my keywords and give me back a whole lot of fucked up shit?
This bugs me so much. It happens constantly. A few days back I searched for a person's name, put it in quotes and I got results with celebrity with somewhat similar name. Zero hits on the person I searched for on the front page. I had to add specifics to the query such as job title to find them.
An easy way to see how a "search engine" has become "vague recommendation engine" is to take a distinctive phrase from one of its results, put it in quotes, and see if it manages to find that page again. Often, it doesn't.
that's an excellent verifiable test! I've been struggling to articulate the behavior too, advertisements and slanted search results have effectively gotten in-between my information retrieval and being steered to products and services. It's so hard to find what I'm looking for a times I often just give up and move onto something else.
Amazon Search is now nearly completely useless for any kind of targeted search. Heaven help you if you're looking for a product without a certain attribute most other products like it all have. There is quite literally no way to filter results against one attribute. Even if Amazon has that product, you won't be able to find it.
I eventually just scripted a separate search engine query that's site specific to Amazon. It works but not as well as it could because it doesn't have access to my purchase history or Amazon's hidden granular category taxonomy.
Oh my god I ran into this yesterday. I wanted a very specific kind of underarmour sweat pants. It gave me every other company competing with underarmour and a bunch of things that are not sweat pants. It’s like they’re not even trying to do “search” any more, but instead just feed your search string into their ad auction system and give you the results. There’s just no way to actually get a specific thing.
I've slammed headfirst into this wall of infuriating frustration dozens of times. Just trying to find a particular kind of LED bulb that has the feature of being dimmable. Any attempt at searching for that term returns all of the bulbs which helpfully mention "Not Dimmable". And there's no way to exclude that string.
It's maddening because Amazon used to have a modern, reasonably capable search function. You could require terms. You could exclude terms. Terms could be phrases. I'm sure they still have all these capabilities, they've just decided to intentionally disable them because their A/B testing indicated that breaking their search would return a fractional percent more revenue by shoveling more unrelated results in front customers. It must work on someone but it's never worked even once on me, because I KNOW what I need and I'm only going to buy exactly that - if I can fucking find it.
I'd actually be okay to let Amazon annoy the NPCs who just clickety-click and buy whatever random shiny shit they shovel in front of them, IF they'd just add something for us technically-minded, engineering type people who are looking for one precise thing only. They can even hide it behind an arcane interface like REGEX. That'll keep the rabble out! :-)
I'm not sure if this is just amazon japan or what, but amazon japan will not only show you things that you didn't search for, it will actively rewrite your search query into a query for those things to gaslight you into thinking you typed it in wrong. And if you try to change the text back it replace it again!
On top of all the other insane choices they made, like removing your search category restrictions if it thinks your query was too precise. I'm close to snapping.
It really feels as though Amazon's greatest fear is the idea that you might search for something and get no results. If we show you BS you explicitly said you aren't looking for then we're giving up the opportunity of tricking you into buying it anyways.
A bit ago I was searching for toothpaste that doesn't have mint in it. This is already a pain at a brick retailer, but I figured Amazon's huge product variety would help. Turns out their search is actively malicious to negative terms because otherwise I could buy just the one thing and be done with my shopping.
I should probably set up a similar homebrew search to get around this. Purchase history is far less important to me because I don't buy much from Amazon.
I haven't had quotations work for me in search in years now, it's really sad how boolean operators have stopped working too. I find it particularly difficult to search for "non latex" products as adding quotes on the total term no longer works and I just get products full of latex. Also I can't use boolean search to find the product because it just ignores the "-" in front of the word and gives a ton of results that match the search term latex.
Just an example where it isn't just making it harder to search for a profit motive but it's actually actively preventing (both Amazon and Google) from showing me the results or even ads for the product I actually want to buy.
If anyone has a good solution to this I would appreciate it, there is often a non-latex version of most all latex based products but finding them online is impossible if you don't already know the brand name!
Google often ignores regular words too, mockingly striking them out. This almost feels like a 1st April joke, not how a search engine is supposed to work.
I'd go out on a limb and guess that the advice is meant for people who don't find them useful.
I don't, for two reasons. First, for looking up facts, they're nowhere near 100% dependable. If I need to check sources, might as well start there. Second, if someone made the effort to put useful content on the internet, I can grace them with a click.
They are useful if you don't care whether the information is correct, don't care where it came from, and don't care that it was stolen from publishers who in many cases invested time and money with the expectation of receiving search traffic in exchange.
I understand the appeal of not wanting to wade through 20 links to find information (especially given SEO stuff that is often on top) but why will people continue to publish as traffic decreases due to AI summaries?
I mean I understand the urgent need to keep up but the problem with "theft" is eventually it drives out honest production and everyone is worse off.
Each paragraph/bullet point typically has a little link icon beside it that directs to 2 or 3 sources. I wouldn't rely on it for anything important but it's OK if you just want a quick rundown on an unfamiliar topic and want to check out likely source material.
I definitely get fed up with AI assistants in the code ide and have said "just shut the fuck up and answer my question". I could just prompt it to be concise but this lets off mor steam.
> Note: Turning off “AI Overviews and more” in Search Labs will not disable all AI Overviews in Search. AI Overviews are part of Google Search like other features, such as knowledge panels, and can’t be turned off
There's no setting for that, Google is intentionally not giving you that choice. You can append &udm=14 to the search URL in browser settings, but that constraints search results to the "web" subcategory, so you also lose fairly useful features, such as the calculator, weather infoboxes, etc.
I work with ML and I am bullish with AI in general; said that, I would pay between 5 to 10 USD a feature or toggle called “No AI” for several services.
For myself I noticed 2 bad effects in my daily usage:
- Search: impossible to reach any original content in the first positions. Almost everything sounds like AIsh. The punctuation, the commas, the semicolon, the narro vocabulary, and the derivative nature of the recent internet pages.
- Discovery: (looking directly to you Spotify and Instagram) here I would add in the “No AI” feature another one “Forget the past…” and then set the time. I personally like to listen some orthogonal genres seasonally. But once that you listen 2 songs in a very spontaneous manner Spotify will recommend that for a long time. I listened out of curiosity some math rock, and the “Discovery Weekly” took 9 weeks to not recommend that anymore.
Google ~13 years ago dealt with a much less complicated search landscape where
1. Its own products weren't (as much) part of the search result quality problem. Today Google has hungry product managers from Ads, Youtube, and various AI products convincing higher-ups that their product deserves higher placement. That placement used to be sacrosanct.
2. The daily volume of AI-generated garbage content 13 years ago was probably a rounding error in today's volume
So Google was operating in a different landscape than Kagi of today. They had to do a lot less to achieve the quality they had.
I disagree that that Kagi "isn't anywhere near as good as Google was ~13 years ago". It's near. For me personally, it's better because I'm never served first-party ads.
Left field tip if you think a search engine is hiding stuff: Yandex. They're not actually Russian any more, but they're far enough down the list of search engines that nobody bothers to DMCA them.
If Kagi made a cheaper "no AI" tier I would be happy to subscribe. AI is costly to run, so even if you don't use the AI it's priced into your subscription fee - you're paying for an expensive product you don't want or use.
e: according to Kagi's pricing page they do have a 'no AI' tier, but it limits your number of searches to 300/month. Seems like a totally arbitrary limitation, but its still better than forced AI.
yeah I hit it in about two. I just signed up monthly to try it out for a bit but am wavering on whether I consider it worth it. It's good but I'm not sure it's enough better than DDG to pay for.
And for AI I'd usually rather have API access and use it with tools rather than a web chat.
I understand your point in general, but I don’t think it applies to Kagi. Users were paying the same monthly fee and then the company added those features on with no extra subscription cost. I also like that it doesn’t clutter my search. It doesn’t appear unless I press the AI button or end my search with a question mark (can also be disabled).
The only extra costs are if you use the (opt-in optional) AI Assistant which is a web UI to access various models for chatting purposes. As an aside, they recently updated this UI so it’s actually usable as a ChatGPT or Claude alternative.
Spotify used to have a "dislike" button for their Discover Weekly which helped with pruning music you don't like, but with the natural law of tech enshitification they removed that feature a month ago.
That was such a frustrating decision. I had almost convinced Spotify that that one time I listened to Lustmord was just a random mood, and I don't actually want to only listen to dronecore for the rest of my life.
I don't know those terms and now I'm afraid to search for them. The Cybernetic Bureaucracy Mind might label me a dissident with terrible taste in music.
I used to always hesitate to use that "dislike" button because I was worried that Spotify would not be able to distinguished between "I will always dislike this song" and "I don't want this song in this specific context"
I insta skipped any song that I liked but didn't want in X context, but disliked songs I didn't want, period, I don't know if it was the intended way, but it seemed to work for me
I can't tell if you misspelled narrow or if "narro" is somehow referring to "narrated" type content we now see so much of. Or even just weird narrative things (eg, recipes).
> I work with ML and I am bullish with AI in general; said that, I would pay between 5 to 10 USD a feature or toggle called “No AI” for several services.
Hard fuck this. I am not giving a company money to un-ruin their service. Just go to a competitor.
I get with a bunch of these hyperscaled businesses it's borderline impossible to entirely escape them, but do what you can. I was an Adobe subscriber for years, and them putting their AI garbage into every app (along with them all steadily getting shittier and shittier to use) finally made me jump ship. I couldn't be happier. Yeah there was pain, there was adjustment period, but we need to cut these fuckers off already. No more eternal subscription fees for mediocre software.
Office is next. This copilot shit is getting worse by the day.
Eh, my partner loves Copilot in Office. There are tasks where it's taking hours or days off the time. Especially web searching and extracting information.
It's also absolutely terrible for image search which has been absolutely poisoned by rampant proliferation of poor quality stable diffusion images - even on stock photo sites.
It got so bad that I had to add a "No AI" flag to my image search app which limits the date range to earlier than 2022. Not a great solution but works in a pinch.
Imagine you are a completely non-technical friend of yours. They are very smart but they do not know a damn thing about configuring computers. They mostly just use their phone and/or tablet.
How much of "just append ?udm=14 to your search query" is absolute gibberish?
Is "install the udm14 plugin" going to make any more sense?
Is "go to udm14.com for all your searches" going to stick? Are there phishing sites at umd14.com, mdm41.com, uwu44.com, and all the other variants they'll probably misremember it as?
"just search for 'fucking whatever' and the AI crap goes away", on the other hand, is funny, uses a common dictionary word that everyone above the age of five knows how to spell, and is intensely memorable.
There's this thing called the internet that can help you find out how to change your default search engine. You might even get some AI suggestions to make it clearer.
You can reset your algo on Spotify! I did and learned a lot. There were maybe 5 songs I wasn’t hearing that I liked, but tens of songs I did not like that I had saved years ago that came back up and were once again swiftly killed by the algorithm after a few instaskips
Yea, it's not perfect, but when I want to find facts (such as about the sizes of nails) and not possible "hallucinations" then it works out. Most of the time I just let the AI overview happen, and I ignore it, so that Google spends more money on a useless and possibly dangerous feature.
Seems like what I do when I encounter an AI answering machine that can answer only simple questions... after "representative" fails, I start saying "Give me a fucking human"
What's likely going on is this. Google is not going to launch an AI computation for every search. That would be lunacy. They're probably identifying some commonly occurring searches, running them through AI, and then injecting that AI result whenever those searches occur. They're probably not exact searches but clusters of similar searches. When your search lands into a topic cluster for which there's a pre-computed AI summary, then you get that summary.
Distracting adjectives like effing probably spoil the match between your search query and the topic cluster.
I found that the "udm" trick disables all "rich content" in search results (ie: stuff that isn't a plain text search result link). But adding the following user rules in AdGuard for Safari does the job:
This is useful info, as Google’s AI summaries have frequently been inaccurate.
Example from yesterday - I was updating my mum’s kindle fire tablet after it had been in a drawer for three years. It was stepping up fire-os versions one at a time and taking hours. So, a quick web search - what’s the latest version of fire os 7?
Google AI confidently answers 7.3.2.9, so cool, it’ll be done soon then. Nope, it kept on going into 7.3.3.x updates.
If you’re going to be confidently wrong, probably best not to try.
One tip I like to give for exploring public data is to do an early search for the word "fuck". It's a pretty ubiquitous word, but one that you assume shouldn't show up for certain fields, so seeing it, or not seeing it, can give useful insight to the the scope of the data universe and collection. Including where/who the data comes from, whether or not any validation exists during the collection process, and how updates/corrections are done to collected data.
For example, you're required to provide accurate info about yourself when donating to a U.S. federal political campaign [0]. Is it possible that someone, somewhere in America is legally named John Fucksalot? Or works for a company named Fucks, Inc? Maybe! We're a huge country with wildly diverse cultural standards and senses of humor. But a John Fucksalot, CEO of Fucks Inc, who lives in Fuck City, Ohio 42069? Probably not, and the fact that this record exists says something about how the rules and laws regarding straw donors are readily enforce. And whether or not an enforcement action happened, what field in the FEC data indicates a revised record?
Seems like this tip can still be useful in the Age of LLMs. Not just for learning about the training data, but also how confident providers are in their models, and how many guardails they've needed to tack on to prevent it from giving unwanted answers.
disappointed that Fucking has self censored, bearing in mind their road safety signs used to suggest local residents were in on the joke https://imgur.com/5KOCwdC
They didn't self censor, they had to change the name because most tourists behaved like assholes. Stealing the road signs, parking anywhere, walking into peoples gardens, etc.
Also the road safety sign, while a funny combination, is pretty standard and can be found all over Austria.
> This is not the first time internet sleuths have discovered a way to disable Google’s AI-powered results. Other methods are more complicated, however, like adding a specific string of characters to the search results page URL. This method of swearing and pleading at Google to “just give me the fucking links” is much more cathartic.
"More complicated" is actually just as simple as going to https://tenbluelinks.org/ and following the instructions. It's so refreshing to just see links when you search, and it's unfortunate that the OP makes it out to be something that's prohibitively difficult to do.
377 comments
[ 2.4 ms ] story [ 295 ms ] threadHmm, I'm not sure about my testing now, even with innocuous stuff the AI thing isn't back. Maybe something I did scared it off.
Will it still work if "fuck" is part of a quoted phrase? If so, you could avoid it by constructing a phrase that contains the term but isn't going to match anything, ex: -"fuck 5823532165".
I eventually just scripted a separate search engine query that's site specific to Amazon. It works but not as well as it could because it doesn't have access to my purchase history or Amazon's hidden granular category taxonomy.
It's maddening because Amazon used to have a modern, reasonably capable search function. You could require terms. You could exclude terms. Terms could be phrases. I'm sure they still have all these capabilities, they've just decided to intentionally disable them because their A/B testing indicated that breaking their search would return a fractional percent more revenue by shoveling more unrelated results in front customers. It must work on someone but it's never worked even once on me, because I KNOW what I need and I'm only going to buy exactly that - if I can fucking find it.
I'd actually be okay to let Amazon annoy the NPCs who just clickety-click and buy whatever random shiny shit they shovel in front of them, IF they'd just add something for us technically-minded, engineering type people who are looking for one precise thing only. They can even hide it behind an arcane interface like REGEX. That'll keep the rabble out! :-)
On top of all the other insane choices they made, like removing your search category restrictions if it thinks your query was too precise. I'm close to snapping.
A bit ago I was searching for toothpaste that doesn't have mint in it. This is already a pain at a brick retailer, but I figured Amazon's huge product variety would help. Turns out their search is actively malicious to negative terms because otherwise I could buy just the one thing and be done with my shopping.
I should probably set up a similar homebrew search to get around this. Purchase history is far less important to me because I don't buy much from Amazon.
Just an example where it isn't just making it harder to search for a profit motive but it's actually actively preventing (both Amazon and Google) from showing me the results or even ads for the product I actually want to buy.
If anyone has a good solution to this I would appreciate it, there is often a non-latex version of most all latex based products but finding them online is impossible if you don't already know the brand name!
Unless your new upstart social service is called SearchEngine+ so you remove it
(Except duckduckgo also seems to semi-ignore it. I'm baffled. I give up. I'm throwing my computer out the window, and moving to the woods)
https://en.wikipedia.org/wiki/Fugging,_Upper_Austria
I'm joking, somewhat, but can we seriously start getting mad about this shit?
You may think you do, but I am certain you do not.
I don't, for two reasons. First, for looking up facts, they're nowhere near 100% dependable. If I need to check sources, might as well start there. Second, if someone made the effort to put useful content on the internet, I can grace them with a click.
I understand the appeal of not wanting to wade through 20 links to find information (especially given SEO stuff that is often on top) but why will people continue to publish as traffic decreases due to AI summaries?
I mean I understand the urgent need to keep up but the problem with "theft" is eventually it drives out honest production and everyone is worse off.
Does anyone know if there’s a uBlock origin filter for this?
https://www.androidauthority.com/how-to-turn-off-ai-overview...
> AI Overview > You can't completely disable Google's AI Overviews in your search results, but you can use workarounds to hide them.
The workarounds didn’t seem to work for me.
https://soitis.dev/ai-overview-hider-for-google
You could also just use the underlying CSS in something like Stylus, or convert it to uBlock filters:
https://github.com/insin/ai-overview-hider-for-google/blob/m...
Can’t remember the last time I used gmail in the browser. And gmail is only ever because startups use gsuite by default.
Google is only going to do more of this shit unless they start hurting from a drop in traffic.
For myself I noticed 2 bad effects in my daily usage:
- Search: impossible to reach any original content in the first positions. Almost everything sounds like AIsh. The punctuation, the commas, the semicolon, the narro vocabulary, and the derivative nature of the recent internet pages.
- Discovery: (looking directly to you Spotify and Instagram) here I would add in the “No AI” feature another one “Forget the past…” and then set the time. I personally like to listen some orthogonal genres seasonally. But once that you listen 2 songs in a very spontaneous manner Spotify will recommend that for a long time. I listened out of curiosity some math rock, and the “Discovery Weekly” took 9 weeks to not recommend that anymore.
TOP 10 X; THE 20 BEST Y; 20 REASONS Z; etc.
I go Kagi and am immediately refreshed.
Yes - a big part of that is that online content is so much worse than it used to be. But it became bad over time because of what Google incentivized.
1. Its own products weren't (as much) part of the search result quality problem. Today Google has hungry product managers from Ads, Youtube, and various AI products convincing higher-ups that their product deserves higher placement. That placement used to be sacrosanct.
2. The daily volume of AI-generated garbage content 13 years ago was probably a rounding error in today's volume
So Google was operating in a different landscape than Kagi of today. They had to do a lot less to achieve the quality they had.
I disagree that that Kagi "isn't anywhere near as good as Google was ~13 years ago". It's near. For me personally, it's better because I'm never served first-party ads.
e: according to Kagi's pricing page they do have a 'no AI' tier, but it limits your number of searches to 300/month. Seems like a totally arbitrary limitation, but its still better than forced AI.
And for AI I'd usually rather have API access and use it with tools rather than a web chat.
The only extra costs are if you use the (opt-in optional) AI Assistant which is a web UI to access various models for chatting purposes. As an aside, they recently updated this UI so it’s actually usable as a ChatGPT or Claude alternative.
> I work with ML and I am bullish with AI in general; said that, I would pay between 5 to 10 USD a feature or toggle called “No AI” for several services.
Hard fuck this. I am not giving a company money to un-ruin their service. Just go to a competitor.
I get with a bunch of these hyperscaled businesses it's borderline impossible to entirely escape them, but do what you can. I was an Adobe subscriber for years, and them putting their AI garbage into every app (along with them all steadily getting shittier and shittier to use) finally made me jump ship. I couldn't be happier. Yeah there was pain, there was adjustment period, but we need to cut these fuckers off already. No more eternal subscription fees for mediocre software.
Office is next. This copilot shit is getting worse by the day.
It got so bad that I had to add a "No AI" flag to my image search app which limits the date range to earlier than 2022. Not a great solution but works in a pinch.
https://github.com/scpedicini/truman-show
If you want something smooth and easy that uses the google engine, visit udm14.com
If you want to integrate the google engine more directly into your browser(s), understand how to use &udm=14
Two different UX's, each appropriate for a different audience.
How much of "just append ?udm=14 to your search query" is absolute gibberish?
Is "install the udm14 plugin" going to make any more sense?
Is "go to udm14.com for all your searches" going to stick? Are there phishing sites at umd14.com, mdm41.com, uwu44.com, and all the other variants they'll probably misremember it as?
"just search for 'fucking whatever' and the AI crap goes away", on the other hand, is funny, uses a common dictionary word that everyone above the age of five knows how to spell, and is intensely memorable.
Yes, the "fucking" trick is awesome too.
Better than Google in every single aspect, except shopping. The shopping results in Google are actually good.
"per capita gdp by country" - no AI
"population of Kansas" - no AI
"lisp interpreter" - no AI
"emacs vs. vim" - AI overview
"size of mit student body" - AI overview
What's going on?
Distracting adjectives like effing probably spoil the match between your search query and the topic cluster.
Safari extension: https://news.ycombinator.com/item?id=41298312
Vs
Single moms in my area fucking
Yep, it works
Example from yesterday - I was updating my mum’s kindle fire tablet after it had been in a drawer for three years. It was stepping up fire-os versions one at a time and taking hours. So, a quick web search - what’s the latest version of fire os 7?
Google AI confidently answers 7.3.2.9, so cool, it’ll be done soon then. Nope, it kept on going into 7.3.3.x updates.
If you’re going to be confidently wrong, probably best not to try.
For example, you're required to provide accurate info about yourself when donating to a U.S. federal political campaign [0]. Is it possible that someone, somewhere in America is legally named John Fucksalot? Or works for a company named Fucks, Inc? Maybe! We're a huge country with wildly diverse cultural standards and senses of humor. But a John Fucksalot, CEO of Fucks Inc, who lives in Fuck City, Ohio 42069? Probably not, and the fact that this record exists says something about how the rules and laws regarding straw donors are readily enforce. And whether or not an enforcement action happened, what field in the FEC data indicates a revised record?
Seems like this tip can still be useful in the Age of LLMs. Not just for learning about the training data, but also how confident providers are in their models, and how many guardails they've needed to tack on to prevent it from giving unwanted answers.
[0] https://www.fec.gov/data/receipts/individual-contributions/?...
Also the road safety sign, while a funny combination, is pretty standard and can be found all over Austria.
Kagi has been a joy to use.
"More complicated" is actually just as simple as going to https://tenbluelinks.org/ and following the instructions. It's so refreshing to just see links when you search, and it's unfortunate that the OP makes it out to be something that's prohibitively difficult to do.