Every recommendation system you typically use from the cloud services (i.e. Youtube, Spotify, Tiktok, Twitter (X), Facebook and so on) is owned by someone whose interests aren't yours. It optimizes for engagement and other private metrics, you can't inspect, can't correct, and can't take with you when the service inevitably dies or even use the same recommendation model for different independent applications. This project aims to solve all of that.

Anagnorisis has the other arrangement that turns the whole client-server architecture upside down. The servers are now pure data-sharing thin hosts and all the main processing, search and recommendations are happening on your machine. You rate local or remote files and other data you own on a scale of 0 to 10. This feedback then used to locally train a recommendation model that scores everything you haven't rated yet automatically. You correct what it got wrong, that correction goes to the next round of training data. You repeat these steps again and again, getting each time model that better and better aligns to your preferences.

The big vision of this project is to provide a platform that creates a local, private model of your interests. That likes what you like and sees importance where you would see it. Then you can use this model to search and filter local and global information on your behalf in a way you would do it yourself but in a much faster and efficient way. Making this platform (in the future) a go to place to see news, recommendations and insights, and so on, tailored specifically for you. As the internet gets populated with bots and AI slop, a platform like this might create a necessary filter to be able to navigate in this chaotic information space effectively.