Almost all of economics optimises the wrong average. Harmoniq is the infrastructure built on the right one — and on everything that follows when you take that fact seriously.
In a multiplicative world — where gains and losses compound — a single trajectory grows more slowly than the crowd, and can drift to ruin while the ensemble average still looks healthy. The gap between them is not measurement noise. It is a property of the process.
When volatility is high enough that μ < σ2/2, your time-average growth turns negative and your path decays to the ruin barrier — even as the expectation grows without bound. Ruin is absorbing: cross it once and no future return averages you back.
Fixed drift μ = 0.05. Push σ up and watch the lone paths fall to the barrier while the gold reserve holds. Raise ρ and watch the reserve lose its power — correlation is what makes pooling fail.
Share the trajectories and you cut the variance that drags the time-average down. Every participant's long-run growth rises — and no expected value is transferred to anyone. Cooperation stops being a moral appeal and becomes the dominant selfish strategy. That is Peters & Adamou's resolution of the cooperation puzzle. CIRES is that pooling operator, built as infrastructure.
The gain is largest for the most volatile, most ruin-prone members — the reserve does the most for those who need it most, without anyone losing. But the equation also names the failure mode. As ρ → 1, Neff → 1 and a thousand correlated members pool no better than one. Diversity is not an ethical garnish on the reserve; it is the quantity that decides whether it works. And because correlation rises under stress, the reserve is engineered to hold ρ low exactly when the barrier is nearest.
If the time-average is what matters and ruin is absorbing, the instruments follow. A model that sees the non-ergodic system, a translator that turns what it sees into incentives, controllers that install the missing brakes, and a reserve that pools.
An agent-based, networked, multiplicative simulator that computes time-average growth, ruin probability and the pooling gain — the instrument that reads critical transitions before they happen.
Turns what the engine finds into live incentives people are paid on — pooled time-average pay, ruin floors, tariffs on degenerative work.
The reserve as a pooling operator on non-ergodic trajectories — proof-of-resilience as ruin-avoidance, diversity as the load-bearing parameter.
The Human Relevance floor as an absorbing barrier the system must stay off, and a tariff that damps the flows which drive correlation toward one.
Directed settlement of multi-capital value, a capacity valve on the flux, and the conserved stock the reserve draws down and refills.
The coordination surface that renders a group to its members, on a sovereign, polycentric rule-set.
A fiduciary AI that acts for the person — on their device, on their side.
The groups that are actually selected for — teams, coalitions, communities — each a pool with a shared buffer.
Democratic middle powers, aligned on capital, money, energy and compute.
Each level aligned to the one above it — the multilevel-selection structure the whole architecture rests on.
Nonlinearity, tipping points, feedback, nesting, path-dependence — the Santa Fe lineage established all of it, and stopped at description. Harmoniq makes three moves past it: the system becomes legible (a non-ergodic world model), controllable (floors and tariffs as installed feedback), and aligned across levels. The science says the economy is a complex system. Harmoniq says it therefore needs a control layer — and builds one.
The reserve's cooperation advantage, formalized for the LML collaboration.
Mandate 1 — the ABM / complexity engine, layer by layer.
The map from the field's canon to Harmoniq's components.