> Many of the features that make YouTube virally successful and a continuing platform are developed based on research produced by its analytics which sacrifice privacy to improve insights into what users care about.
I think you're overselling it. YouTube is successful predominantly because they originally had first mover advantage and now have a direct link from google.com.
The problem decentralized services have is the same problem Google has had in taking over for Facebook. You need to provide something which is enough better than the incumbent that it can overcome the switching cost. The problem has nothing specifically to do with decentralization, and time has proven again and again that it can happen, every time one network replaces another. The next time it happens we just need the new network to be a decentralized one.
> To make a comparison - which are you betting on in 2016 - IRC or Slack?
Here's a different question -- which are you betting on still being used in two decades, IRC or Slack?
That's one of the huge advantages of decentralization. Things stick around because they don't die when one company goes out of business. Which means the process of incremental improvement can refine them into something that, once it has the incumbency, is almost impossible to displace. Examples: TCP, email, Unix/Linux, DNS. None of them are perfect but they'll be with us for the rest of our lives.
> Even if you set up a successful open source project to replicate YouTube, how will you develop it as well when your best employees are poached because they aren't paid as well as centralized companies?
The same way it works for Linux or BSD.
> How about the analytics features? How do you design an anonymous or decentralized recommendation engine with the same or similar utility? How do you power search relevance in a meaningfully private and decentralized way? How do you add the social networking features that people care about without a centralized social network?
Decentralized search and recommendations are hard. So do the easy part first.
Look at YouTube. It's really two independent pieces. One is a web host for videos, the other is a player app with search and recommendations. So you decentralize the hosting and then people make apps with central servers that do recommendations and search. Tomorrow somebody else will figure out how to decentralize the other part. You can make progress without having to solve every problem in the same place at the same time.
> The basis of modern product research and improvement is data tracking and analysis.
The underlying assumption is that you can't ever be finished with anything. But the more time passes the more refined the product becomes and the less you need to fiddle with it anymore.