Excited to see this is planned! I've been experimenting with book recommendations using Hardcover data and wanted to share some findings that might be useful.

I built a collaborative filtering system on top of the Hardcover API using BPR (Bayesian Personalized Ranking) matrix factorization. A few things I learned that might help inform the official implementation:

1. BPR loss over pointwise loss — Hardcover ratings skew heavily positive (~95% are 3+ stars). BPR treats recommendation as a ranking problem rather than classification, which handles this imbalance much better. In my tests it doubled precision.

2. Filtering out ultra-popular books matters a lot — Dropping books read by >10% of users from the training data doubled the "hit rate ratio" (how much better than random the recommendations are). Everyone has read Harry Potter — that's noise, not signal.

3. IDF-weighted shared books — When showing why two users are matched, weighting by inverse popularity surfaces the meaningful connections. "You both loved this obscure novel" is more compelling than "you both read a bestseller."

4. Implicit feedback signals are rich — Using granular weights (rating/5.0 for rated books, 0.7 for read-without-rating, 0.3 for want-to-read, 0.0 for DNF) captures much more signal than binary liked/didn't-like.

I have a working prototype deployed at:

https://book-friend-finder-i2nrrpteiq-uc.a.run.app

The code is open source at:

https://github.com/RaggedR/book-friend-finder.

Happy to share more details or collaborate if it's helpful to Adam's work on this!

I would love to see a readers also enjoyed/similar books section on each book's entry page in addition to general book recommendations based on users' activity.

Would be lovely if recommendations were created on the basis of any of want to read/reading/have read.

can you make it that way so that any member can add recommendation under each book also give the vote system so that the most relevent book come up in recomendation in top here is a example https://comick.dev/comic/the-thousand-faces-actor