Bluesky is very useful to store information on users' existing accounts.
I'm currently building a review system for my open source Web map https://cartes.app, based on Bluesky. Not trivial though, you have to create a lexicon and maintain a DB based on the Bluesky stream.
You can go pretty far without your own DB. Depends on the types of queries you need to make. For my project[1], I was able to use getRecord[2] for a lot of the data that needed fetching on the client-side.
Except that it got around 2.9K reviews by now, which is more then you have right now. Furthermore, we shouldn't further fragment the few open source review efforts we have.
Many OSM apps will also be reluctant to adopt a closed source solution that might be closed of any moment. And under what licenses will those reviews be? As MapComplete developer, I can not and will not be adopting a system based on Bluesky
The idea of fragmentation depends on a model of decentralization that still makes the platform inseparable from the data it works with. AT separates concerns so that the priorities of your data host and the priorities of your platform host can conflict without one being able to control the other even if you don't self-host. All the reviews are just an entry in your PDS, so it's all there for any new or existing platform.
Thousands of people have already set up their own PDSes and it's inevitable managed hosts will appear soon. Blacksky just started migrating people over to its own PDS. AT's credible exit is close to reality after about two years while all the promise of ActivityPub and predecessor protocols has yet to materialize after over 15 years.
I'm currently building a review system for my open source Web map https://cartes.app, based on Bluesky. Not trivial though, you have to create a lexicon and maintain a DB based on the Bluesky stream.