Nostr transmits notes and other stuff over relays. This is good. However developers are using nostr as a database. While, yes, you get a sugar-rush by doing so, this is non-optimal for several reasons:

- the content is only one field which is a text string, that is not easy to work with

- meta data is added to the content, but it is often misused

- the query languages is incredibly basic, worse that what existed in the 70s

- the database indexes are permissioned and centralized, it can take a year or to get approval to use an index

- a small number of developers get approval almost immediatly, the vast majority of innovation is kicked into the long grass

- the meta data violates basic computer science ideas of Entity-Attribute-value and has hard to work with array lists

- it is well-known for decades that connection-oriented networks dont scale

- the kinds (ie types) are rationed out by a central authority and no more than 100 are alloed

- entries such as profile are centralized, for example you are allowed one website, to do something like adding a 2nd website would be a multi-year project

- follower counts have to be added manually, so if you have a large number of follows then its impractical to make a query

This is not bad, because is shows that nostr can improve in many ways. But the first step is to acknowledge the limitations, before any of it can be fixed. Improving the devX would allow nostr to grow many OOMs.