The models

Models you can run, not stories you can tell

A theory you can’t execute is a narrative. The lab builds composable, categorical models where the double-entry accounting invariant is enforced by the data structure itself — one specification mapped to many coherent semantics, so the same model yields a simulation, a steady state, and a calibration.

The stack

How it’s built

Godley tables
The double-entry stock-flow grid that is Minsky’s signature, expressed as a composable categorical object.
Catlab / StockFlow
A Julia engine doing the composition, vector-field generation, simulation and calibration.
Browser-native
A no-install, collaborative front end over the categorical model; the solving stays on the server.
Multi-semantics
One spec mapped functorially to a vector field, a steady-state system, a stochastic version, and a calibration model.

Method

Reproduce, couple, audit

Reproduce
Replicate the published result exactly — the Keen 1995 replication as the acid test.
Couple
Add the money, selection and exergy layers Minsky alone doesn’t carry.
Audit
Machine-checkable models where the invariants are enforced by structure, not by the interface.