this post was submitted on 28 Jul 2026
96 points (91.4% liked)

Privacy

10453 readers
106 users here now

A community for Lemmy users interested in privacy

Rules:

  1. Be civil
  2. No spam posting
  3. Keep posts on-topic
  4. No trolling

founded 3 years ago
MODERATORS
 

I've spent the past two years building NeoSearch with a simple premise: search got bad not because the web got worse, but because the incentives did. Ads, profiling, and SEO all pull results away from what I actually wanted.

NeoSearch is my attempt to fix that. It's private by design (no cookies, no IP logging, no profiles, no ads), and it uses AI to improve results rather than replace them: independent sources get ranked above affiliate listicles, titles and snippets are rewritten to say what a page actually is, duplicate results get merged into one enriched entry, and "Lenses" let you flip between meanings — search "eagles" and choose the band or the football team.

Currently it takes the top results from Google's index as raw material, then re-ranks, rewrites, and enriches them so heavily that Google's #20 result can land at #1. That's stage one, to prove the experience.

Stage two is crawling the parts of the web mainstream engines ignore.

The end goal is an open, distributed search index anyone can build their own engine on. I don't want to be the next Google, I want search to stop needing one.

As of this week the whole thing is open source (Apache 2.0) on GitHub, so you can audit it or run your own.

It's free, I'm funding it personally, and there will never be ads. I'd genuinely love your feedback — especially the searches where it does worse than Google. Those are the ones I learn from. Ask me anything!

you are viewing a single comment's thread
view the rest of the comments
[–] Imhotep@feddit.fr 2 points 4 days ago

It's fast, polished, really quite impressive. Maybe too AI-centric for my taste, and yet I like how the results are presented, and they were quite relevant at times... and bad too. For example I was searching for chatons.org (software) and it only gave me results for kittens* (chatons in french). While recognizing I was in France but giving all answers in English.

I wonder though, if I want to search something simple, say I forgot the exact URL of a website, how much ressources are being wasted explaining what the website does instead of simply giving me a link?

I like how Kagi does it, when I need a short AI answer I add a questionmark to my query.

I'll keep testing this, it looks very promising. Best of luck to you!

*And the AI saying kittens are hairless at birth? I don't use google so I assume this is the type of answers it gives, man that's terrible