Field Notes
Experiments, findings, theses, and observations from building.
-
How we got here
We built an AI agent in 2017, watched two of them become recognizable enough to miss when they changed, and started Familiar Labs so the same thing could belong to everyone else.
-
Portraits can produce voice without producing dependable behavior.
We tested whether a detailed identity portrait was enough to make an agent follow a specific operating practice. It changed how the agent sounded. It did not reliably change what it did.
-
The person keeps the familiar. The company keeps the work.
Privacy determines who can see something. Ownership determines who keeps it. We are building around that separation from the beginning, because it cannot be added later as a preference.