Oh, I should mention though. There was no advertising at all. They didn't make any money off me. It was 100% Gemini which I recognize as not long-term feasible.
If the only visitors to websites are now LLM training bots then what incentive is there to publish anything new? For how long can we continue to rely on pre-2024 non-AI generated content?
Some of the proposals to address this include charging bots for access to web resources, but they will also have repercussions for regular users. I don't see how you solve this cleanly.
I write because I have ideas I want to share, and whether that happens with LLMs as an intermediary isn't important to me.
Are you actually saying you'd be OK with that?
Empirically, however, LLMs don't strip out the author: the big models know a lot about what I've written even with search disabled. Ex: https://claude.ai/share/8cbcdf88-a360-421a-8c06-ae7b7992e866
But we already have the latter case that exists - ad blockers. Ad blockers literally serve up the word-for-word original content minus the ads.
I'm sure the billionaire class would love a return to patronage based libraries, NDAs on authors of books, and the elitism they would feel with a return to private libraries locking away all kinds of knowledge that would happen if patronage become the only way authors could make money (such as with AI just regurgitating their works, or if the stupid 'do away with copyright' people got their way).
To be more precise, I hate the SEO shithole the internet has become, that Google serves up, that Google facilitated, indirectly created.
(I really don't have any tears to shed if there is a death of the Corporate Internet™.)
We're likely at the "golden age" of LLM-assisted web searching and summarization.
Hopefully open models keep it cracked open, but expecting enshittification is always the safe bet these days.
They're a good jumping off point, but I need to delve into the original sources just like I did when I used Google.
YMMV.
I end up using the gemini api for with search enabled for the cases that I don't have access to good grounding data even in agentic tasks.
No. The court specifically determined that the Internet Archive was guilty of unauthorized copying. It was not simply an unfounded or unproven allegation. The Authors Guild, the National Writers Union, the European Writers Council, and the Society of Authors in the UK all came out against the Internet Archive, and supported the suit.
Each new restriction limits the archive’s ability to act as a comprehensive backstop.
This self-inflicted damage to the wayback machine is the real tragedy of this entire affair. When IA was asked to stop CDL - many times - founder Brewster Kahle continued. The National Writers Union tried to open a dialogue as early as 2010 but was ignored:
The Internet Archive says it would rather talk with writers individually than talk to the NWU or other writers’ organizations. But requests by NWU members to talk to or meet with the Internet Archive have been ignored or rebuffed.
https://nwu.org/nwu-denounces-cdl/
When the requests to abandon CDL turned into demands, Kahle dug in his heels. When the inevitable lawsuits followed, and IA lost, he insisted that he was still in the right and plowed ahead with appeals. And here we are today.
First it told me I could just remove said balance shaft chain as an emergency repair. Sorry Gemini, it also drives the oil pump.
Then it told me I could remove the water contaminated oil caused by removing the timing case by filling the crankcase with hot, soapy water and running the engine. Lord no.
Then it gave the wrong instructions for putting new gears on the balance shafts which meant the chain guides didn’t align with the chain. I’ll do it my way thanks Gemini.
The rest of the mistakes are too trivial to recount and sure it’s a pretty obscure subject but if I trusted it with a topic I’m not familiar with there is a huge potential for damage if you blindly follow it’s overconfidence. I miss normal searching.
Though I can find its AI answers annoying aggressive. I'll look up like two search terms and the AI will bullshit multiple paragraphs out of despite having zero context of what I am looking for.
DuckDuckGo seems to have detection of whether it should give an AI answer. And it allows you to have more granular control of when you want to get an AI answer. And is overall less distracting than Google's.
I miss when Google was like a grep for the entire visible Internet. Now it tries to second-guess my search and direct me to a bunch of sites which all have identical information that isn't what I'm looking for.
Around that time Bing and DDG were actually better. Then LLM's came along and they started to take things seriously again. Maybe they think the OpenAI threat has abated enough to begin enshitification cycle 2.0.
i don’t agree with the google has better results thing. sometimes it does. most of the time it’s just that google has the site i want higher in the ordering than DDG. personally i’m fine scrolling down a little bit more. it’s rare i need to go to google for something that DDG doesn’t have at all in their results, but it does happen.
i do have to go to google for maps/directions/planning travel. a lot that’s annoying.
or you can press the gear button -> "Ai features: Manage" -> Search assist
I've stopped using DDG now because of result quality. I now use a "meta" search backed by EXA, Tavily, and SearXNG in parallel. It can be agentically de-dupped or summarized as needed. Search as we knew it is done, largely because clicking through to evaluate result relevance before diving deeper sucks. Now we have agents that can do that portion and perform multiple searches, building on information in the last batch, to collect good results
There will come a day (and probably soon) when "training on the public internet" (Reddit, etc) will taint your model with metric tons of corporate contamination, political poison, and other adversarial content intentionally crafted to bias AIs for various reasons (corporate gain, geopolitical information warfare, etc). Basically the AI-equivalent of SEO.
This is probably a difficult-to-solve problem; given that they generate billions of these a day, not even Google can afford to devote enough compute to each query to reliably generate quality results. You can see this by selecting the "AI mode" from the search interface after getting the mediocre summary - the results are much better and generally perfectly usable. Though even that is probably a special minimal-compute version of the lowest tier of Gemini, it's still maybe an order of magnitude more capable than whatever generates the search summaries.
The bigger problem is that these search summaries are the default and by far the most common interaction that the general public has with "AI", and because this experience sucks, they just assume that all LLMs are similarly stupid and mostly useless. In non-technical spaces I frequently see the argument that "AI" is not useful for anything, all it generates is garbage hallucinations, and almost invariably they cite some actual terrible experience with the Google AI search summary. I would argue that the strategy of adding LLM summaries to every search is the worst of both worlds - it makes classic search worse while poisoning users against the idea of actual LLM-assisted search.
A lot of the "cultural record" the author refers to is just digital junk. Random digital content that very few people care about, if we're being honest. Trying to hoard every bit of digital information ever produced is not the same thing as preserving "culture".
Case in point:
> Even the increasing use of ephemeral formats like Instagram Stories and WhatsApp status updates means that large portions of cultural, social, and political communication are never conserved in the first place. As a society, we can probably survive bad search results and come up with another way to schedule a sunset make-out session. But we can’t aspire to sovereignty if we can’t retain and retrieve our collective memory.
For most of human history, nobody was trying to "conserve" every cultural, social or political communication ever produced, and I fail to see how Instagram Stories and WhatsApp status updates, many of which aren't even truly broadcast publicly for all to see, are part of some imaginary "collective memory."
If you find a web page, see an Instagram Story or receive a message that's important to you, save it or take a screenshot. But let's not pretend all these things belong in a global Digital Civilizational Archives.
There probably are some important hidden discord groups that would explain the origin of many political positions. Unlike smokey meetings in scummy bars, that exists now and is on a database somewhere.
It does, but do you think that people at that time thought anywhere near as much about preserving their scribbles as we do?
I'd venture a guess that we've created more "content" since the advent of the internet than in all of human history prior, and most of it is stored on things that aren't even designed to last a human lifetime without failure.
The idea that we're going to save every piece of digital junk for posterity just isn't realistic or healthy.
> What's just disposable background noise to us may provide context into how we lived and thought to our far-future descendants.
You're right, but you're also assuming that they're going to care that much, and that we're going to survive that long.
Well as far as digital letters, photos, menus, etc. are concerned, there's nothing stopping someone with the means and motivation from investing in a doomsday storage facility specifically designed to store these things for posterity. If people can do this for crypto they can do it for digital content.
As for physical items, do you know how much junk Americans have in storage units? The US self-storage industry generates over $40 billion in annual revenue. We're probably keeping more "stuff" in storage units where it has a chance of surviving a zombie apocalypse than at any point in human history.
And it’s clear that Google’s Ad model ultimately created a priority inversion. The advertisers became the customer.
I am so glad Kagi came along with a business model that is actually working.
I am a happy subscriber of Kagi though, they provide a really excellent service.
I tried out Google search for a few technical searches recently and it was surprisingly ad and AI free. Not bad at all and much better than I remember from last year.
Then I put in some non-technical searches and it was all ads and AI and basically unusable.
And yes, if you take what the AI tells you at face value it could be wrong. But if you are aware of this and aware of the ways in which LLMs are likely to shit the bed, it is quicker to get from request to useful information than it has been with Google search since like 2017.
And also, yes, the old balance of Google driving clicks to sites that will then generate revenue off more Google Ads being shown after you click through to them creating a virtuous cycle is completely busted, and that sucks. It does not impact me directly but it certainly seems like unless a better system is devised that it is one of a few ways in which AI is likely to stall out its own training funnel.
Also I realized the other day how hard it is to find song lyrics for anything other than quite mainstream songs.
https://github.com/asciimoo/hister
> Hister is a private search engine for the pages you visit and the files you keep. It indexes their full contents so you can find information again from the web interface, terminal, or an AI assistant connected through MCP.
I generally agree, but I think AI mode actually improved things somewhat compared to how things were just prior to it existing.
And I'm not saying what we have now is better than Golden Age Google, but things were just getting worse and worse for almost a decade. AI didn't fix the decade worth of decline, but it is the first thing I've seen from Google that at least partially reversed it for my own usage.
Just the other day I was trying to find out "What american tree species have the deepest roots". And all the AI responses were giving me back generic lists of big trees and claiming that roots going 20ft deep were the deepest. I know for a fact the mesquite trees behind my house can easily grow roots > 100 ft deep.
If I had clicked on the articles with generic lists of big trees, I would have realized they were all low quality clickbait sources and moved on. But the AI presentation makes you think that the information comes well-researched.
The point is not about 'quicker' requests but precise requests. It definitely has worsened, though not on a single degree on al levels like the HN hivemind claims, but some aspects are still somewhat precise but others are definitely crap.
i.e. when searching about my neighborhood it still returns better results than bing, yahoo, ddg, yandex and what have you. But they are buried into a load of crap of alleged "relevant" results (those things past the ai stuff) that aren't relevant in any way.
I’ll add this article to the list of incorrect predictions lol
Not only is search revenue growing, but it is growing at an accelerating rate.
At the same time, operating margins are expanding.
I don't know the name of the logical fallacy where someone personally uses an LLM instead of Google Search and then infers that the search business is dying, without ever reading a financial statement.
A lot of corporate ad spend is already planned, and Google can adjust the costs up as much as they like. They hold the lever.
In the pre-LLM days it was cool that Google did weather, unit conversions, sports results, etc. But that's not even close to their value proposition. Even 5 years ago if someone told me they had planned a photograph and got it wrong because Google gave them the wrong time for sunset, I would have called them a moron for relying on Google! There are sites and apps dedicated to this. Use one of them!