Show HN: 13M opinions on 48k products extracted from Reddit and HN (trustmebro.reviews)

Resubmitting to give this another shot.

I made this a couple of months ago. Extractions were done with Gemma 4 26B MoE running on pairs of NVLinked 3090 cards with vLLM, in my basement. Seasoned with a Claude vibe-coded frontend.

The travel opinions on a map are pretty neat: https://www.trustmebro.reviews/c/travel?view=map

If someone wants to take it over, you can mail me.

9 comments

[ 7.1 ms ] story [ 307 ms ] thread

garageguideapp 14h ago ↗

I love this idea. I will research some products and come back with results.

davidzweig 12h ago ↗

For me it's quite interesting to see lists of products ppl geek out about and discuss, categories I wasn't really aware of.

Theentity 13h ago ↗

Such great work !!

davidzweig 12h ago ↗

Thanks!

headsman771 13h ago ↗

You should share the data, because this is definitely being shut down by Reddit. Making money on "their data" is definitively not fair use. Good luck though.

davidzweig 12h ago ↗

The data is out there, not very difficult to find. It's 3-4 weeks of work on a box with 8 3090 cards, less if you rent some B200s etc. It's not monetized currently.

davidzweig 12h ago ↗

I'll just give a bit more context:

Last year I used a similar approach to extract Podcast recommendations for language learners: https://www.languagereactor.com/c/es/pd/t_pd_mix_es

Then realized it generalized well to products: https://www.languagereactor.com/c/es/rs/t_rs_all_es

I had it on my to-do list for a while to make a general site for recommendations. The latest models are now able to coordinate and manage the various moving parts.

sama004 6h ago ↗

damn thats some crazy data work

davidzweig 5h ago ↗

I got a GB200 slice (4x B200) in the mail. :)