Servers and server farms comprised of PCs are horribly complex to scale in a reliable way, and the programming models force programmers to have to either reason about complex "distributed computing" problems, or externalize state into a scalable "database" that has already thought about the problem.
End-user PCs act as Frankenstein contraptions, both containing private data for some applications, and containing no data for others. In the former case, the OS ABIs and application paradigms do not treat data security and privacy as a first-class concern, so ransomware and malware abounds. In the latter case, our PCs are brain-dead for most daily productivity without being tethered to a "cloud" (aka Mainframe) run by companies whose sole investor remit is to monetize our data and/or our interactions with other people via their free applications.
A revolution in computing is in the offing: one that is centered around users, their data, and their interactions with each other. Information is the privileged and important asset; not the APIs of the hardware that runs the compute. The topology of human interaction and human groups around that information system is far, far more important than the hardware network topology, which should be an irrelevant detail of routing.
Every output an input.
Which researchers or companies are working on this architecture, other than the negative examples of socialmedia/adtech, WeChat and China’s social credit system?
Honorable mentions for tackling part of the problem go to IPFS and Scuttlebutt.
There's a lot of people who understand the larger vision around the decentralization crowd if you can find them in the haystack of hypesters.