Why we are building Entropy
Every major strategic decision is a bet on human behavior.
Whether you are restructuring a corporation, launching a disruptive product, or shaping public policy, you are effectively wagering on how thousands of independent actors will react to a new reality. The stakes are absolute, but the tools we use to place these bets are archaic.
We rely on surveys, which capture what people say they will do, not what they actually do. We rely on focus groups, which reduce complex populations to a handful of voices in a conference room. We rely on historical data, forcing us to drive into the future while looking in the rearview mirror.
The fundamental problem with traditional research is the observer effect: You cannot test the future without creating it.
To measure employee sentiment regarding a reorganization, you must ask them about it—triggering the very anxiety you sought to measure. To test a pricing strategy, you must expose it to the market—tipping your hand to competitors. To gauge voter reaction to a controversial policy, you must announce it—and live with the consequences if you guessed wrong.
We are building Entropy to break this paradox.
Simulation as Strategy
Entropy is a simulation engine that allows leaders to test the future in a low-stakes environment before committing to it in a high-stakes reality.
We do not ask people what they would do. We create synthetic populations—statistically grounded in real-world demographics, psychographics, and network structures—and we let them act.
This is an ontological shift in how we understand the world. We don’t ask clouds if they intend to rain; we model atmospheric dynamics. We don’t ask bridges if they intend to collapse; we model structural physics. Yet, when it comes to the most complex system of all—human behavior—we are still stuck asking for opinions.
Entropy brings the rigor of computational modeling to social science.
-
Grounded Fidelity: We generate agents based on deep research, not generic templates. A simulated German surgeon possesses the specific professional attributes, institutional rank, and peer incentives that drive real surgical behavior. A simulated swing-state voter possesses the specific media diet and economic anxieties of their district.
-
Dynamic Evolution: Surveys capture a static snapshot. Simulation captures motion. We model how information travels through a network, how early adopters influence the late majority, and how sentiment decays or hardens over time.
-
Convergence: As we increase the number of agents, variance decreases. By simulating at scale—ten thousand agents, one hundred thousand agents—individual noise cancels out, and population-level truths emerge.
The End of Guessing
The implications of this technology are profound.
An organization can simulate a restructuring plan to identify specific pockets of retention risk before a single memo is sent. A product team can model the churn impact of a price increase across different user cohorts without risking a public backlash. A campaign can test the second-order effects of a policy announcement, measuring not just immediate reaction, but how the narrative evolves over weeks of social transmission.
This provides the one thing traditional research cannot: The ability to be wrong in private.
You can run the simulation a thousand times. You can fail nine hundred times. You can adjust the variables, refine the messaging, and iterate on the strategy until the model converges on success. And only then do you execute in the real world.
We model supply chains before we build factories. We model aerodynamics before we fly planes. It is time we modeled the human response before we make history.
The future of strategy is not polling. It is prediction.
Entropy is currently under active development at this time of writing. The following posts in this series explore the technical architecture enabling this vision.