From Volunteer Computing to Application-Driven Crowdcomputing

Volunteer computing settled one half of a hard problem: people will contribute their machines to work they find meaningful, and BOINC proved it at the scale of millions of machines. It never settled the other half. Generating work still means operating a computing project, so a supply of computation demonstrated at planetary scale remains usable by a handful of institutions. This commentary argues that the next step is a shift in where work originates, from dedicated projects to the domain applications people already use, and examines the model that follows: initiators inside everyday software, reviewed controllers defining each kind of computation, and one shared network underneath. It states the limits of that model plainly, in result validation and in the boundary of coarse-grained parallelism, and asks what kind of computer the arrangement points toward. Twenty-five years of volunteer computing have settled a question that once sounded utopian: will ordinary people, in large numbers, donate their computers' idle time to work they find meaningful? They will. Millions of machines have passed through projects built on BOINC, the Berkeley Open Infrastructure for Network Computing, since SETI@home made the idea famous, and the science is real: pulsars discovered, protein structures predicted, climate ensembles run. SETI@home itself stopped distributing work in 2020, which says something about the durability of the model rather than against it: the platform and the community outlived the project that started them. What volunteer computing has not solved is the other half of the problem: where work comes from. In the established model, to generate work is to operate a computing project. You run a server. You define workunits. You recruit a volunteer community and keep it engaged, sometimes for decades. The supply of computation has been demonstrated at planetary scale; the ability to ask for it has remained the privilege of institutions. This essay argues that the next step for the field is a shift in where work originates: from dedicated projects to the domain applications people already use. I will call the wider model crowdcomputing, by analogy with crowdfunding: people contribute computation, rather than money, to projects they care about. The variant I examine is application-driven crowdcomputing, in which work is generated from inside those applications, and the person generating it never has to think like a project operator at all.