Quinn Campbell says the future of on-chain games depends on building a system that can fail fast—and learn faster.