Show HN: I built a 2B-page search engine, independent of Google/Bing
5 Chief_Searcha 7 7/17/2025, 4:45:47 PM
Hi HN,
For the last 18 months, I've been working solo on building a completely independent search engine from scratch. Today, I'm opening it up for beta testing and would love to get your feedback.
The project powers two public sites from the same 2-billion-page index:
Searcha.Page: A session-aware search engine that uses a persistent browser key (not a cookie) for better context.
Seek.Ninja: A 100% stateless, privacy-first version with no identifiers at all.
The entire stack is self-hosted on a single ~$4k bare-metal EPYC server in my laundry room (no cloud, no VC funding). The search pipeline is a hybrid model, using a traditional lexical index for the heavy lifting and lightweight LLMs for specific tasks like query expansion and re-ranking. It's an experiment in capital efficiency and digital sovereignty—proving you don't need Big Tech APIs to compete.
I’m looking for feedback on search result relevance, speed, and the clarity of the privacy models. Please try it out and let me know what you think.
Links:
https://searcha.page
https://seek.ninja
Thanks,
Ryan
The one thing that I would like to see is the "More results from <site>," links for sites with a lot of results. :site works so not a deal breaker, just a would be nice.
The first results were good for the search I did. However, I went back and searched for something else entirely, and I got results similar to the first search. That seems bad, since they were unrelated searches.
Now I can't search anything because it's timing out.
Did you crawl yourself or using common crawl?