Live data from Hacker News

Reddit can't build a better search engine

ruky.me

121–130 of 289 posts

Re: Reddit can't build a better search engine

#121
post #103

i remember history very differently. the old search engines before google analyzed only page content in isolation, which lead to SEO sites appending invisible blocks of keywords at the bottom or putting loads of barely related keywords in the page title. google was the one that managed to break this cycle with pagerank. suddenly, the spamword filled trash sites were worthless because nobody (of worth) linked to them.…

This.

In the early days of Google search, there were lots of other search engines (Hotbot and Altavista come to mind), and they were crap compared to Google.

Re: Reddit can't build a better search engine

#122
post #92

This thesis is off. It's not as if Reddit just appeared on the internet yesterday. It's been a place people look for this kind of info for a long time, and people have been trying to game it for just as long. And it already has a good search engine - that just happens to be Google. So it still leaves the question: why does Reddit still have such a high signal-to-noise ratio. A few things I've noticed: 1. A lot of con…

Recently have noticed a business creating fictional threads on reddit (multiple on different communities) working as an AD to sell something. Example: „What is the cheapest best cooling for cpu now?” All discussions point to product that more experienced users know its crap. But still the wrong product is at the top with most (botted) upvotes. Companies realized that ppl look for opinions on reddit -> that created de…

Thank you for sharing this view. There are a few nails in reddit's coffin and this is certainly one of them.

I used to be able to gauge real world opinions on reddit pretty accurately, i.e. be aware what was happening before the curve. That is gone now.

Re: Reddit can't build a better search engine

#123
post #92

This thesis is off. It's not as if Reddit just appeared on the internet yesterday. It's been a place people look for this kind of info for a long time, and people have been trying to game it for just as long. And it already has a good search engine - that just happens to be Google. So it still leaves the question: why does Reddit still have such a high signal-to-noise ratio. A few things I've noticed: 1. A lot of con…

Recently have noticed a business creating fictional threads on reddit (multiple on different communities) working as an AD to sell something. Example: „What is the cheapest best cooling for cpu now?” All discussions point to product that more experienced users know its crap. But still the wrong product is at the top with most (botted) upvotes. Companies realized that ppl look for opinions on reddit -> that created de…

This problem will only get worse as language models get better, making it cheaper to flood a forum with bot comments that look more and more like real humans. Moderators and users alike will spend far more time and resources weeding that stuff out, while it takes far fewer resources to post that stuff in large volumes.

The only way to counter that is to have some kind of account verification requiring 1 account per person. Such schemes are all either very unpopular ("Why does reddit need my passport photo!") or very ineffective ("So I just need a new IP and I can sign up for a new account? Great, I'll set up 1000 accounts")..

Re: Reddit can't build a better search engine

#124

>So how can we fix the system? The best way to fix the system is to prioritize websites that are there to share knowledge, not websites with their primary priority to make ad revenue. Different solution: - Make a search engine that is actually good. Blacklist all SEO spam such as pinterest, allow complex queries, restrict by date, site, regex, etc - Charge 1 cent per search, or as much as is necessary Why has nobody…

>Why has nobody done this? What am I missing?

The answer is in your proposal:

>- Charge 1 cent per search, or as much as is necessary

90% of people do potentially hundreds of searches everyday. Especially non expert users that practically use the search engine to search for the domain of the sites they need to go too.

The people ready to pay a fixed fee /per search wouldn't be able to cover even a fraction of the infrastructure costs of : "search engine that is actually good"

Then there is the question of what good means.

Google's solution that "outside links" are considered as "votes" that this site is good is what made it good but also what made it vulnerable to SEO.

Do you have any other idea on how do you decide if a result is good or not beyond the literal text search?

Re: Reddit can't build a better search engine

#125
post #92

This thesis is off. It's not as if Reddit just appeared on the internet yesterday. It's been a place people look for this kind of info for a long time, and people have been trying to game it for just as long. And it already has a good search engine - that just happens to be Google. So it still leaves the question: why does Reddit still have such a high signal-to-noise ratio. A few things I've noticed: 1. A lot of con…

Recently have noticed a business creating fictional threads on reddit (multiple on different communities) working as an AD to sell something. Example: „What is the cheapest best cooling for cpu now?” All discussions point to product that more experienced users know its crap. But still the wrong product is at the top with most (botted) upvotes. Companies realized that ppl look for opinions on reddit -> that created de…

Yes, Reddit is being gamed. The question as with any system is: To what extend and how does it compare to alternative systems?

The capital that is most relevant on Reddit is status. It is very easy to have people participate in Reddit, which makes it inviting enough to start building status but it's somewhat harder to actually do. Users care about their status and about their communities and Reddit combines the two in a nifty way, that already gives provides some slid resistance against gamification.

Sure, you can infiltrate the system with any/or a lot of accounts, careful VPNing, lucky/clever reposting and also keep the guise up (you must, since your portfolio of both content and accounts is under constant review, as long as it remains visible). To get a highish ranking article on google I mostly have to write what feels like convoluted infomercials (or maybe let GPT3 do it for me). There is an industry of people being paid to do so. All it takes is one mediocre writer and some time.

To me, the former system seems very brittle and expensive to game, the later much less so, specially since the prevalence of convoluted writing for SEO purposes make the distinction between SEO bullshit and expansive content increasingly difficult: People are getting used to informercial type content on the web and so it is increasingly what they learn to expect. And we give the people what they apparently want.

Re: Reddit can't build a better search engine

#126
Reddit would need strong anti-fraud/anti-spam expertise, because currently most of BlackHat SEO masters are focused on Google.

Subreddits with their own moderators can work for the current traffic (and probably 10x that), but the quality of the platform is likely to degrade if reddit received >10% overall google search traffic.

I am not saying Reddit can't do that, and I would be really happy to see it happen.

Re: Reddit can't build a better search engine

#127
> So how can we fix the system? The best way to fix the system is to prioritize websites that are there to share knowledge, not websites with their primary priority to make ad revenue.

Which makes it a shame that Google is in the business of pushing ad revenue rather than search.

Re: Reddit can't build a better search engine

#128
> I’d still prefer a targeted ad over a random ad that is unrelated to me.

No, a million times no. A targeted ad requires my personal information being collected, and the resulting assumption of my needs may be offensive and even dangerous.

I prefer a random ad. Or better, a place where I can just search the product I want in detail.

Re: Reddit can't build a better search engine

#129
post #128

> I’d still prefer a targeted ad over a random ad that is unrelated to me. No, a million times no. A targeted ad requires my personal information being collected, and the resulting assumption of my needs may be offensive and even dangerous. I prefer a random ad. Or better, a place where I can just search the product I want in detail.

A targeted ad can be a contextual ad, with no personal information involved.

Re: Reddit can't build a better search engine

#130
post #12

But the practice of appending 'reddit' to your Google search isn't depending on Reddit building a better search engine. It's using Google's search engine, and trying to direct it to a database that you hope has the answer you're looking for. So there's that. Of course, if that practice became widespread, then people could still try to stuff Reddit posts with junk in the hope that it gets picked up by Google in these…

There was a post, which trended on HN yesterday (2000+ upvotes) that stated Reddit is sitting on a gold mine and they can build a better search engine because people are appending 'reddit' at the end of it. This is a response to the original post, explaining why the logic behind is it false.

@rukshn Thanks for your post. Personally I'm not convinced by the idea of adding site:reddit.com to your Google search query either, not just because it will get gamed by SEO practitioners as you suggest, but also because there's a lot more to the internet than reddit and a lot of great content hidden away which isn't on reddit. I also believe that a more general solution to better search is both to reward informational sites and also to penalise advertising sites - that's why I built a search engine with a PageRank-style algorthim which also detects result pages containing adverts and heavily downranks them.
Post reply on HN