Science
How do you build a synthetic human?
Most synthetic-human systems start from a description of a person. We think the more interesting question is what happens when you start from their behavior.
Imagine trying to create a digital copy of your best friend. You could start with what you know about them: their age, where they live, what they like, what they believe, what they tend to buy, what they are afraid of. Then you hand all of it to an AI and say "pretend you are this person". That is roughly how many synthetic-human systems work today, and it runs into a simple problem: knowing facts about someone is not the same as understanding them.
The slideshow below walks through the architecture we are testing at Synthetic Users, one layer at a time: a Core of the things about a person that change slowly, the memories a situation brings to mind, the temporary state they happen to be in today, and only then the reasoning that produces an answer.
None of this is assumed to be an improvement. Every version is tested against real human responses, and any layer that doesn't make the Synthetic User predict the organic user better has no reason to stay in the system.