SCIENCE

This isn't speculation. It's built on peer-reviewed research from Stanford, DeepMind, and decades of social science.

KEY RESEARCH
STANFORD 2023
Park et al.

Generative Agents: Interactive Simulacra of Human Behavior

25 AI agents living in a simulated town. They formed relationships, spread information, coordinated activities — all without explicit programming. Emergent behavior from memory and reflection.

25 agents·Emergent social behavior
STANFORD/DEEPMIND 2024
Argyle et al.

Out of One, Many: Using Language Models to Simulate Human Samples

1,052 agents calibrated to real survey data. Individual responses matched human answers with 85% accuracy. Not just aggregate patterns — individual-level prediction.

1,052 agents·85% individual accuracy
THEORETICAL FOUNDATIONS

Built on proven models.

Our agents incorporate decades of social science research on how people actually behave, decide, and influence each other.

ASCH CONFORMITY

Social pressure shapes belief

People change stated opinions to match groups, even when they know the group is wrong. Our agents model this conformity pressure.

GRANOVETTER WEAK TIES

Acquaintances spread ideas

Novel information travels through weak social connections, not close friends. Our network topology reflects this.

ROGERS DIFFUSION

Innovation adoption curves

Innovators → Early adopters → Early majority → Late majority → Laggards. Our agents have varying adoption thresholds.

KAHNEMAN SYSTEM 1/2

Two modes of thinking

Fast intuitive judgments vs. slow deliberate analysis. Our agents switch between modes based on stakes and cognitive load.

LIMITATIONS

What we can't do.

Honesty about limitations is how you know to trust what we can do.

Not a crystal ball— We show likely friction points, not guaranteed outcomes. The future has variance.
Calibration depends on data— Agent accuracy reflects input quality. Garbage in, garbage out still applies.
Black swan events— We can't predict truly unprecedented events. Neither can anyone else.
Cultural specificity— Agents trained on one population may not generalize to others. Context matters.
Not a replacement for reality— Simulation informs decisions. It doesn't make them. You still need to ship.

Want to know the story behind this?

Read the origin story
Join Waitlist