The killer app for AI might just be unenshittifying search for a couple of years.
Then SEO will catch up and we'll have spam again, but now we'll be paying by the token for it. Probably right around the time hallucination drops off enough to have made this viable.
Only the corporate web which values quantity over quality. Have you tried Marginalia Search? It's refreshing, although it doesn't index enough stuff to find what you're looking for, most times.
"Anthropic's Super Bowl ad humorously criticized OpenAI's decision to introduce ads to ChatGPT, featuring a scenario where a man seeking advice is interrupted by an unexpected advertisement." - ddg search assist
Too much money in ads, and search is just a huge cash pipeline straight towards it. No way we can have non-ad-infested llm search out in the wild from any major vendor in upcoming future. Google-fu just becomes llm-google-fu, while sometimes it goes off rails and then apologizes in that typical super annoying way (and screws up something else).
Maybe smaller ones can somehow provide almost comparable but ad-free service, heck even mildly worse but genuine results would win many people over, this one included.
2012-2014 ish. I had a job which often involved searching for specific part numbers, and at some point during that job quotes stopped giving me exact results. They’d give me something close but incorrect. Like “ABC123” would show me “ABC456”. The real part numbers in question were much longer, in the range of 20-30 chars, so sometimes it was hard to notice that the search had ignored my quotes at first.
They might have fixed it in more recent years, but to me that was when the tide started shifting in the mentality behind google search as a product/service
The thread seems to be about the opposite problem. The OP can't find the page they're looking for because Google is too strict about whitespace, according to the top comment.
Google gave me more than 2 pages of results, while Bing gave me only 1 page.
After that, both Google and Bing provided countless pages full of results about Hoodoo, Voodoo, Zoodoo and the like.
The impossibility of making exact searches is what annoys me most in modern search engines.
I might make a typo sometimes, but I would prefer to correct myself when that happens, instead of the search engine always assuming that I am a moron that cannot type, thus offering every time "helpful" corrections.
I tried to repeat your experiment. I entered "xoodoo", surrounded by quotation marks into the Google search bar. I got 29 pages of results. From what I can tell by the previews, every result matches the string "xoodoo" and not any other similar string. Are you sure you are using the exact search functionality and not just typing the word into the search bar?
Yeah, if you consider a military-grade AI/LLM with access to all military info sources, able to analyze them all much quicker than a human… there’s no way this isn’t already either in progress or in use today.
Probably only a matter of time until there’s a Snowden-esque leak saying AI is responsible for drone assassinations against targets selected by AI itself.
>Yeah, if you consider a military-grade AI/LLM with access to all military info sources, able to analyze them all much quicker than a human… there’s no way this isn’t already either in progress or in use today.
Still wouldn't mean much. Wars are won on capacity, logistics (the practical side, not ability to calculate them), land/etc advantages, and when it comes to boots on the ground, courage, knowledge of the place, local population support, etc. Not "analyzing info sources" at scale which is mostly a racket that pretends to be important.
Slightly off topic, but: is it just me or does it become uninteresting to read things like this when the whole process is basically “so then I asked Claude this, and then I asked it this, and then this…”
I guess it takes away the intrigue of the project because anyone could do this (ask ai), and the only thing human left about it is the creativity of the idea itself. There’s not much merit to the effort.
Edit: nothing against the author or anything, it’s fun to do projects like this.
But I always kinda likened AI output to your kids’ artwork - to you it’s the best. To someone else, it doesn’t have as much impact.
I agree. I thought this was really interesting until I got to the point of them needing to fire up an assistant to write basic Apache config - ok fine maybe they were just introducing the assistant - but the next paragraph they were talking about using multiple agents for this despite not knowing why. I gave up at that point, it's not even slightly interesting for me
Why would implementation details be led by product? “Undo” is an action that the user may want, which would be led by product. Not the implementation in the db.
I believe that was the point. Soft delete isn't a product requirement, it's an implementation detail, so product teams should talk about the user experience using language like "delete" or "archive" or "undo" or "customer support retrieves deleted data".
Yeah: You don't "delete" a bank account, you close it, and you don't "undo", you reopen it, etc. The processes have conditions, audit rules, attached information, side-effects, etc. In some cases the same entity can't be restored, and you have to instead create a successor.
"Undo" may work as shorthand for "whatever the best reversing actions happen to be", but as any system grows it stops being simple.
reply