Draft:Macrocosm Group

• Comment: In accordance with the Wikimedia Foundation's Terms of Use, I disclose that I have been paid by my employer for my contributions to this article. Giabharadwaj (talk) 15:53, 18 June 2026 (UTC)

---

Macrocosm Group

Type	Private
Founded	2023
Key People	J. Doyne Farmer (Founder & Chief Scientist) Dan Eichelsdoerfer (CEO) Eric Beinhocker (Board Member) David Young (Board Member)
Headquarters	New York, New York
Website	macrocosm.group

Macrocosm is an American and UK company founded as a spinout from Oxford University in 2023 by complex systems scientist J. Doyne Farmer. The company uses techniques from complexity economics and agent-based modeling to enable scenario analysis of macroeconomic shocks, investment strategies, and policy decisions.

Macrocosm describes its longer-term aim as building an Economic World Model, a simulator intended to serve as a "physics engine" for the economy. The company positions this work within an emerging category it calls "complexity intelligence." As a Public Benefit Corporation, Macrocosm aims to "support better decisions by leaders in business, finance, government, and civil society to create an economy that is more prosperous, sustainable, and equitable."

History

Origins at INET Oxford

Macrocosm’s conception is closely tied to the work of its Chief Scientist and founder, J. Doyne Farmer, who has spent his career applying physics-based computational methods to economic problems. Farmer conducted research in complexity economics at the Santa Fe Institute for 13 years before moving to Oxford, where he is the Baillie Gifford Professor of Complex Systems Science and the Director of the Complexity Economics program at the Institute for New Economic Thinking (INET). His work at Oxford INET helped lay the intellectual foundations for Macrocosm, especially research in two areas: the economic impacts of COVID-19 and forecasting technology costs.

During the COVID-19 pandemic, Farmer and his colleagues used an agent-based simulation of production networks to model the economic implications of lockdown and reopening scenarios in the United Kingdom. The team predicted the economic impact of COVID-19 lockdowns at 21.5% GDP loss; compared to the actual 22.1% GDP loss, the model outperformed both The Bank of England and consensus forecasts at the time. According to Farmer, the COVID-19 simulation reflects the ability of agent-based modeling to address questions that traditional tools "cannot even ask."

In a 2022 paper published in Joule, the Oxford team applied their technology cost forecasting approach to a probabilistic assessment of the global energy transition. Their research showed that a rapid green energy transition could yield net savings relative to a fossil-fuel-dependent baseline by 2050. The paper also noted that traditional energy models have consistently overestimated the future costs of renewable technologies. Farmer has described the failure of conventional economic models to anticipate the pace of the renewable energy transition as a central motivation for founding Macrocosm.

Technical approach

Complexity economics

Complexity economics is the theoretical framework underpinning Macrocosm’s approach. Originating through the Santa Fe Institute in the late 1980s, complexity economics understands the economy as a complex adaptive system rather than a system in equilibrium. Complexity economics helps solve several fundamental problems with mainstream economic models, including the assumptions of perfectly rational agents and economic equilibrium. This approach is much closer to the real world, where agents make decisions based on simple rules (often derived from behavioral economics research) and the economy is constantly shifting and evolving. The New York Times describes Farmer’s approach as applying "insights from chaos theory and complexity economics" to macroeconomic problems that conventional models have struggled to address.

Agent-based modeling

Macrocosm’s simulations are primarily built on agent-based modeling (ABM), a bottom-up computational approach rooted in complexity economics. In ABM, individual economic entities are represented as autonomous agents that operate according to behavioral constraints. This process allows agents to be heterogeneous and adaptive, generating macro-level patterns as emergent properties of individual interactions.

Unlike traditional models, Macrocosm’s approach extends beyond neoclassical assumptions that each agent is perfectly rational and fully informed. Every agent in a given market is individually represented and makes realistic decisions that change as the broader economy changes. The company’s simulations are designed to address counterfactual and novel scenarios where conventional models, trained on historical data, are liable to fail.

Applications

Supply chain resilience

One of Macrocosm's primary application areas is supply chain shocks and resilience. In a 2025 paper presented at the EurIPS Workshop on Differentiable Systems and Scientific Machine Learning, Macrocosm’s research team introduced a differentiable supply-chain agent-based model implemented in JAX. Running simulations on GPUs, the team showed that large production-network ABMs could be calibrated far faster than previously possible.

In a 2026 paper, the team further explored the conditions under which supply chain networks shift from resilient to fragile states. They found that competitive pressure on firms to minimize inventories pushes production networks to a critical point at which small, localized shocks can lead to large aggregate disruptions. Macrocosm has also published a paper examining the macroeconomic consequences of a potential closure of the Strait of Hormuz. The paper applies the company’s supply-chain modeling framework to a real geopolitical risk scenario.