The project was founded in 2014. It was heavily focused on building a federated protocol before there was anything related to end usability. Maybe they've just flubbed the implementation and done a poor job of it, I don't know. I can tell you from the external perspective it's not doing a great job at anything technical or end UX.

WhatsApp nailed distribution, Signal does privacy well, Telegram is simple and scalable. What is Matrix? None of the above. I'm sorry, that might come across as harsh but I see a lot of wishful posts here that kind of throw users in that direction without real insight and knowledge into it. We don't want DIY that's endlessly complex and hard to maintain. The reason centralised apps work is because there's no setup, so really you're fighting frictionless experiences. Want to make self hosting a better experience, make it frictionless.

I'm a fan of Matrix but am inclined to agree. In 2018 I looked at the Riot (Element) codebase with a hope to contribute to improving the client. In my estimation, the interfaces and data structures were bonkers convoluted and the maintainers gave dire warnings about how all of them were subject to radical change. I put my attention elsewhere.

I wish they'd focused on building a messenger first, rather than trying to be a Slack/Teams/Discord. That said, it's taken a while but the Element client is dramatically better now. It still doesn't have the polish of other messengers, but I'm rooting for them to dial that in.

Matrix does decentralization/federation well? The problem is that it's a philosophical/political goal and most people place other values (social cachet, ease of use, total cost) above that.

Another problem is that a focus on federation limits your appeal to people who like to self-host things.