Equilibrium economics answers the wrong question. It asks what price clears the market when everyone is rational, identical, and finished adjusting. That is a useful cartoon for a seminar, and a dangerous one for a trading floor or a supply chain.
The alternative is not to abandon rigor — it is to model behavior that is genuinely adaptive. Traders who learn. Firms that imitate. Shocks that propagate through network topology rather than dissipating smoothly across representative agents. The methods are now mature: agent-based models, calibrated networks, non-linear stochastic dynamics. The bottleneck is practitioners who can translate between the research and a real business problem.
That is the practice.
Stress-test pricing, launch strategy, or trading rules against synthetic populations of heterogeneous agents before committing capital. Finds the failure modes averages can't see.
Map the graph — who actually depends on whom, two and three hops deep. Identify nodes whose failure triggers cascades disproportionate to their size or spend.
Conventional VaR underestimates regime change. Complexity-based indicators — correlation structure, critical slowing-down — give early warning of phase transitions conventional models miss by construction.
Apply network theory to deal sourcing, synergy modeling, and post-merger integration risk. Synergies live in the graph of who-talks-to-whom; so do the integration failures.
A working toy of the Santa Fe Artificial Stock Market. Three populations of traders interact: fundamentalists pull price toward a latent value, chartists chase the trend, noise traders add liquidity and randomness.
Raise the chartist fraction — watch bubbles, crashes, and fat tails emerge from the same rule-set that used to converge quietly.
NO EQUILIBRIUM ASSUMED.
NO REPRESENTATIVE AGENT.
The practice was founded on a simple wager: that the methods of complexity economics — agent-based simulation, network analysis, non-linear dynamics — are ready to leave the journals and solve real business problems. Most organizations still rely on models built on equilibrium assumptions they would never use to run the rest of their operations.
Current work spans correlation-structure analysis for equity portfolios, network models of supply-chain and financial exposure, and agent-based simulations of market microstructure. All output is Python-native, reproducible, and versioned publicly where possible.
If you have an interesting question — inside a firm, a research group, or a hiring pipeline — the fastest path is a direct note. Short is fine.
Complexity Insights