But whether or not this form of mind model can do justice to the person’s true personality while they were still alive is up for debate. He calls Roman 2.0 “a compressed JPEG of a mind”; many details are missing, but the broad structure is strangely intact. Originally, he had been developing sideloading primarily as a way to build his own “mind model,” a kind of personal synthetic double that could outlast him. In this light, Roman 2.0 is a test case in a long-term philosophical engineering project. “You read only AI-generated answers… without editing.” To make his point, he describes having placed Roman 2.0 into a group conversation with ten people.