Social Physics Lab is a math-centric research team building the protocols behind human and agent coordination. Hyperstition is its enterprise practice.

The role blends research and engineering. You design experiments on trust scaling, blame assignment, and negotiation between model agents. You build the environments, verifiers, and evaluation harnesses they run on, and diagnose why a training run does or does not improve. What holds up ships against a client's live data.

We are looking for people who have done exceptional work in mathematics or at the frontier of modern AI. Public work counts for more than credentials, and degrees are not required.

If this is you, write to shivi@socialphysics.inc with links to the strongest thing you have built or proved.