I think the ideal place this would add value is pre-launch of a feature or MVP. With enough randomization and "thought" patterns to test out your features it c

I think the ideal place this would add value is pre-launch of a feature or MVP. With enough randomization and "thought" patterns to test out your features it could be a very novel approach to testing that differs from existing load testing & integration testing. Randomization is the key though, the synthetic users would need to be capable of rooting out those random edge cases that you can only find by typically using a product "incorrectly". Cool idea though!