If Martin is correct, that effectively outlines a product spec for open-source, distributed compute where broad swaths of groups, individuals, and organizations contribute compute resources to training runs for large-parameter open-source models