A couple of ways from my understanding. We have different focuses in our UX and UI as we, for example, feature reviews directly next to the product and show products 1 at a time instead of a listing view. We also place more emphasis on having a semantic search where you learn about the products being offered and how they're relevant to your specific situation instead of a keyword based search. From a business standpoint, we're also affiliates of Amazon and Stylevana while Looria isn't.
Interesting, haven't seen one of those yet. With public video+audio models getting much better that will only get way worse over time. Excited to see what YT/Google decides to do about it.
Example: search for "MSI G27C4X" on YouTube, for me at least both the first and second results are fake robot reviews. There are a couple of real impressions videos by real people but for some reason YouTube sorts them below the AI spam. One of those spam channels is posting multiple reviews per hour, with >11,000 videos and counting.
Great question. This would be a bigger issue if we were only aggregating results and summarizing them, but because we both aggregate and show (in our opinion) the highest credibility reviews from YouTubers (and other sources like blogs once we add them), our idea is that while the general mass opinion can be shifted through campaigns like that, the top end of the spectrum should hopefully still remain pure.
If on the other hand the top end of the spectrum is corrupted, then hopefully the masses can compensate for that. If both are corrupted and all of the data sources available are, then it really comes down to our ability to filter out LLM or promoted content which comes down to how well they can hide it. AI detection tools have been scaling alongside models, so it's also a question if that will continue over time. We'll think of some more advanced things if that becomes a bigger issue for us :)
At the end of the day, if a company can do a coordinated advertising campaign across the internet over months to block out any negative opinion, it's a big deal for both us and the social media/data sources we pull from that's going to be a challenge we have to deal with.
The top comment on that clip scores home that point. While there's no way to verify that, I wouldn't be surprised if it's a lot more than we realize. That being said and to play devils advocate - if all of them are already astroturfed with some people discovering that, then how much do most people care about it?
What's happening here is that we put your search query into a model that tries to figure out what skincare product characteristics (e.g. Vitamin C, Retinol, etc) would be good for you, but when it sees something that isn't a skincare product or a skincare concern, it gets confused and doesn't return anything.
This is something I ran into as well, mainly bc of your title and description. I tried it before reading your whole post, so I didn't know it's only skincare for now.
> Launch HN: Lumona (YC W24) – Product search based on Reddit and YouTube reviews
> Hey HN! We are Lumona (https://lumona.ai), a product search engine
Thanks! We just watched your review together in the living room, and we really appreciate your thoughts+detailed feedback. The list of items is an interesting idea that we'll think about how to fit into the ux. Comparisons is definitely something we want to add later down the line as well.
The idea of restaurants, like you mentioned, would be really great to have. It's not an immediate priority, but once we get Tiktok/short form videos on the site and integrate it well, it'd be really exciting to make and use.
That's what I've heard as well. Also, the lag to when you actually get paid is super painful.
Also curious - how do you think about affiliates as someone who runs an ecommerce site? Are there any reservations about whether services like us take search traffic or ads revenue?
Thing is, there aren't many marketing firms that don't have (or can't buy on a work-for-hire basis) upvote/downvote networks. It's trivial to promote comments in ways that look organic. It's equally trivial to downvote commercially harmful comments into oblivion.
What's more, Reddit posts are usually actively discussed only for a day or two. But votes can be cast, and new comments can be added, for months. One strategy is to wait until the conversation has completely died down, then hijack it with new comments that somehow seem to get as many votes as they need to rise to the top. When, a year later, somebody digs up that thread on Google, they'll see the promoted comments first.
Reddit has severe structural flaws that are, I think, unfixable. In making the upvote/downvote thing a kind of game, and in enabling easy throwaway accounts whose votes are weighed in the same way as those of the longstanding accounts of regular commenters, they've naturally made their forum easy for commercial interests to game.
Great points, not sure those are fixable either. It definitely has some advantages from those same things in some respects, but monetizing reddit outside of ads (i.e. ecommerce like tiktok/instagram) because of these things is going to be a challenge for sure.
