AI is coming for our jobs. What will we do about it?
We're building an open-source economic simulator to model what happens when AI displaces workers on a large scale. Our research points to three possible futures with AI.
This is such an unprecedented situation that it can be hard to get one's head around it. Here are core concepts that are crucial to understanding what's coming.
AI is very close to being able to replace you at work
Today's best AI already surpasses most employees in general knowledge and reasoning. The only things still missing — persistent memory and full-screen interaction — don't require major breakthroughs and are coming soon. If your job runs on a screen and keyboard, the AI that can replace you is either already here or months away.
If companies can replace you, they must replace you
Under capitalism, replacing workers with AI isn't optional — it's mandatory. Companies that move first win market share, boost margins, and see stock prices soar; the ones that hesitate fall behind, lose investors, and go under. No CEO can choose to keep paying humans when their competitors aren't.
AI will create new jobs. And AI will do those jobs.
Every past technological revolution created new jobs and even whole new industries that needed human workers. People assume AI will follow the same pattern. It won't, because the technology that destroys the jobs is the same technology that fills the new ones.
Layoffs drive layoffs: The downward demand-employment doom spiral
The downward demand-employment spiral is one of the best-understood mechanisms of modern capitalist economies, but it is not well understood by the public because the government has stepped in to prevent these from occuring for more than a century. When laid off workers cut their spending, other businesses lose revenue and therefore cut staff. Those cuts in turn drive more cuts. When the spiral begins, business owners across the economy begin to anticipate a general downturn and the cuts accelerate even more. Therefore, if 5%, 10% or 25% of jobs are lost to AI automation, that will drive job losses on an even greater scale.
You can't tax enough to replace the lost wages
"Tax the robots and give everyone a check" — that's the plan. But run the numbers and it falls apart: replacing the full incomes of displaced workers would cost many trillions per year, while the tax base itself is shrinking as AI replaces the high-wage workers who fund it and competition drives profits toward zero. Even the most expansive UBI imaginable would be a fraction of what you need to replace lost demand.
Even if you could tax enough, UBI is a perpetual motion machine (and those don't work)
Set aside the practical impossibility for a moment. Even in the fantasy scenario where the government can collect enough taxes, UBI under private ownership still fails — because every cycle, the owners keep a cut as profit, and the machine winds down. The government must print money to replace what leaks out, or tax at 100%, which is just public ownership with extra steps.
We're building a tool to think this through
The debate about AI and jobs is mostly hot takes and hype. We need real models to help us understand how AI layoffs might actually play out, what their impact will be across the economy, and how various policy responses could work in practice. So we're building one: BigSim, an open-source agent-based economic simulator documented following the ODD protocol framework. It's a work in progress — you can follow our development, review our entity classes, state variables, behavioral rules, and process schedules, or contribute code and ideas on GitHub or Discord. We want this to be a community resource for understanding the coming economic transformation.
BigSim is designed to run at up to full 1:1 scale — approximately 130 million households and millions of firms and organizations, each following behavioral rules under bounded rationality. All transactions are stock-flow consistent. The model is being calibrated to the US economy with 376 exogenous parameters, runs on a monthly time step over a 40-year horizon, and features sectoral heterogeneity across five entity classes. No assumptions about equilibrium. Just the dynamics playing out.
This is a political question, not a technical one.
The only solution currently being offered in public discourse to mass AI layoffs is some version of UBI — a check from the government funded by taxing the machines. But that arrangement leaves ownership of the most productive technology in history in very few hands, with everyone else dependent on transfers. That's a choice, not an inevitability.
There are other paths. Public and shared ownership models could distribute the gains more broadly — the way we already do with roads, utilities, and public universities.
The question is who decides: the people building the AI, or the rest of us?