I don't think anything built on traditional tech will catch on. Whenever I see anything touting decentralization, I always look at the setup instructions becau

I don't think anything built on traditional tech will catch on. Whenever I see anything touting decentralization, I always look at the setup instructions because I think the setup complexity will dictate the adoption potential. With Mastodon, it looks like Ruby on Rails, Node.js, PostgreSQL, and Redis for the backend stuff. For me, the only realistic way I'd run a node would be via Docker (Compose) which means I'm limited to something non-redundant (ie: no HA) that only scales vertically, at least easily. > The article mentions an admin of 80K users spending 300+euro/mo I think it's well understood that centralized systems win because normal, non-technical people don't want to deal with the complexity of managing their own servers and that devolves into the situation like the one you pointed out where normal users consolidate on a single node and the owner of that node ends up stuck with all the cost, moderation, maintenance, etc. The thing that I think isn't considered enough, or at all, is that decentralization of control doesn't need to mean decentralization of infrastructure. Imagine something like Mastodon but built with newer tech that makes it deployable to Cloudflare, Netlify, Vercel, etc. Think along the lines of using something like SvelteKit + adapters. I can see all of those infrastructure platforms getting to the point where a custom deployment could be distilled down to something as simple as installing an app on a phone where the only thing you need to do is feed it a domain. As long as it's federated using a custom domain, making it simple enough for a normal person to run an instance increases the decentralization of control which is more important than decentralizing the infrastructure in my opinion. I think one of the main things that we need to make that more realistic is for someone like Cloudflare to provide a local runtime for their core services (Pages, Workers, Functions, KV, R2, etc.). Having something vertically scalable is adequate because that's no worse than what most of us get with a Docker based deployment which is the maximum complexity most people will be willing to deal with, at least the people I know. The other thing that would help a lot would be if those companies I mentioned collaborated to have deployment compatibility. IE: I should be able to write one app and deploy it to any one of them as long as I use a standard set of features. I think that would be amazing from the perspective of being a customer / dev, but I don't know how realistic it is on the business side of things.