Show HN: Got sick and tired of AI in search engines, so I made my own
glass.8ball.spaceI got sick and tired of how much AI was being pushed at me from nearly every search engine. So I built my own, Project Glass.
It is built on a few core beliefs: Transparency and Privacy is paramount. AI isn't required. Real user engagement is the best signal.
It has a clear transparent (puns intended) algorithm, that you can view on the home page, it takes the following things into account for the ranking: 1. Title 2. Snippet 3. Number of clicks/Click Rate 4. Recency (with a slow decay)
It uses 0 front end Javascript, and no AI in any part of the creation or use of this.
I have been working on this for a while, and this has been my daily use search engine for a few weeks now, if you can please try it out, let me know what you think, and please be kind!
Ill be around all day to answer questions and review feedback.
Thank you! I have pushed quite a few more updates to this since this post:
Added Wikipedia snippets
Refined the algorithm
Changed how stuff gets pulled and crawled. I’m here to answer any questions or feedback anyone has if they want to share! I just use Google with AI summaries disabled What index does it use? Is it your own? It uses my own database and if it can’t find enough results it will pull them from both google and bing and pulls the results to the DB, but I also have my own crawler that is out as well. That’s pretty cool. How many pages do you have in your database so far? Also since you are working on a search engine, I would recommend reading this article: https://archive.org/details/search-timeline Last I checked it had just over 9 thousand, I think it was like 9076 or something like that.
And thank you for the read! That was pretty interesting! No problem. You might consider using data from the Common Crawl to boost your index size. If you get the extracted text files (called WET instead of WARC), they don’t take up much space. I have one from 2014 that has about 73’000 pages in it, and it only takes up about 300mb uncompressed. Those files are surprisingly easy and fun to work with, and downloading them will probably always be faster than crawling on your own. If you use files from the older crawls it will probably make your product more distinctive, but there are probably a lot of 404’s so you might have to give people an option to view the cached page or go to the Wayback Machine. You probably don’t have the resources for this, but I would love it if someone made a search engine that lets you search though all 115 or so crawls that they have, which would be around 100 billion pages and take up around 816 TB. thank you! Ill look into that a bit, i don't have the space for the full thing yet but some day!