1.) Well, both startups and enterprises. both can benefit from reduced costs.
2.) Haven't worked out any figure yet.
3.) I was thinking an agent could run on the laptop/machine exposing compute and making sure it doesn't sleep, also report availability i.e if it can schedule a job or not.
4.) Correct, it should be a drop-in replacement type of thing, where for instance only changing the runs-on bit in github actions should suffice. Machines could became part of a node-pool via vpn? and jobs could be scheduled on the ones that are available.