What should nations do when AI can replace most jobs?
We're building a simulation to show how different policy responses might play out.
BigSim is an open-source agent-based economic simulator. It models interactions of millions of households, firms, banks, and government institutions using behavior grounded in observed real-world dynamics rather than frictionless textbook assumptions. It can be used to test policy responses to AI job displacement and other shocks.
Why we're building this
Simulations don't predict the future. They help policymakers, leaders, and social movements understand their options better.
Provide full salary replacement for all laid-off workers? Tax companies to raise that revenue? Can that actually work under capitalism? How would other options work, such as a public ownership model? Get a rich picture of how these choices might actually work in the real world.
We're simulating the economy with over 400 parameters and 16 interacting entity types so we can move beyond slogans to plans grounded in how the economy actually behaves — including frictions, stickiness, institutional quirks, and feedback loops that standard models often smooth away.
How it works
Choose your assumptions
Set up scenarios about what could happen — and how government, corporations, and social movements could respond.
Test policy responses
Try out different interventions and see how they affect the entire economy — employment, wages, housing, debt, and more.
Explore the results
See how scenarios play out in simulation to better understand possible limitations and opportunities in our future.
Unlike many macro policy models that rely on representative-agent structures for tractability, BigSim tracks every individual. Macro outcomes — unemployment, GDP, inflation, inequality — emerge from millions of micro decisions under real-world constraints. Key institutional behavior (including central-bank reaction regimes) is configurable, so you can compare historical dynamics with counterfactual policy paths.
Where we are now
BigSim is in active development. Here's what exists and what we're working on:
16 entity types — households, firms, banks, government, markets, and more with real BLS/Census/FDIC data
420+ parameters documented with sources, ranges, and audit status
59 economic flows mapped between entities
52 rules defining agent behavior across 10 simulation phases
89 historical regime events (1986–2024) for calibration
12 counterfactual and AI displacement scenarios
Julia simulation engine — agent-based, monthly time steps
Declarative rule system — define model behavior in YAML, not code
Interactive simulator — download and run scenarios on your own hardware
How you can help
The hardest part of building an economic simulator isn't the code — it's getting the parameters and behavioral rules right. We document assumptions with sources so collaborators can audit, challenge, and improve them. We need contributors who will hunt inaccurate representations of behaviors, flows, policies, and other aspects of the model.
Review parameters
Browse the registry and check if our assumptions match reality. Flag anything that looks wrong.
Browse ParametersSuggest improvements
Missing an entity type? Think a flow is wrong? Have better data sources? Open an issue or join the discussion.
GitHub IssuesJoin the community
Talk with us on Discord. Ask questions, share ideas, or just follow along as we build.
Join DiscordBigSim is a project of New Consensus, part of the Mission for America.