Found something good?

We actually had a very similar concept before a year then Darkbloom, and it was live and working for some time.

The biggest challenge was generating enough demand consistently. Even when node operators received workloads, many would turn off their machines because of heat, electricity costs, or simply because they needed to use their devices themselves.

That created a reliability problem. If someone turns off their machine in the middle of a job, the user's workload gets interrupted or lost. For businesses that rely on stable inference, that is a major issue.

The other thing people often miss is the economics. The platform has to make money, and the node operator also expects to be paid. Once you account for platform margins, payouts, electricity costs, hardware wear and tear, and periods with no workload, the numbers usually look much less attractive than the calculator suggests.

That is one of the reasons many companies eventually moved toward using data centers and professional compute providers instead of relying on individual laptops and desktops. The infra is more reliable, easier to manage, and can handle larger customers who need consistent uptime.

So I would treat the "$423/month" figure as a best-case estimate rather than something you should expect in real-world conditions. The real question is whether there is enough sustained demand and whether providers can keep machines online reliably enough to support it.