Society, in silico.

Closed, aggregate, unvalidated models drive the financial decisions that affect every household. The Congressional Budget Office scores trillion-dollar bills with models no one outside can inspect. Financial institutions price risk using proprietary simulations they can’t share. Researchers publish with data they can’t replicate.

The result: society-wide miscoordination. Not because people disagree on values, but because they can’t agree on facts.

There is no open alternative for economy-wide, household-level simulation of tax and benefit policy. Existing tools are either aggregate (CBO), closed-source (proprietary), or narrowly scoped (individual calculators). No system combines validated statute encoding, synthetic microdata, and population-scale simulation in an open, queryable platform.

Cosilico builds the shared substrate: open-source statute encodings validated against multiple external systems, synthetic populations calibrated to reality, and APIs that make the whole simulation queryable. We encode law as code — every formula traced to its statutory source — and build commercial APIs on top of the open infrastructure.

We use AI to accelerate statute encoding, with a multi-stage validation pipeline: syntax validation, reference checking, test-driven development against statute text, and external validation against TAXSIM, PolicyEngine, and other systems. The AI proposes, the pipeline validates.