Flow-based carrying capacity beats a stock buffer for clean resource-grounded growth

When modeling emergent population/settlement growth, my first cut PRODUCED the need-resource into a persistent stock and grew if fed. Bug: the stock buffer let population OVERSHOOT on an accumulated hoard, so final size reflected history/starting-conditions, not the resource endowment (two very different endowments converged to the exact same size). Fix: compute carrying capacity from the FLOW — K = (labour*land yield + imports)/need — and grow logistically toward K with no persistent granary (drain/spoil surplus so capacity tracks fresh supply). That gave clean monotonic resource-grounding and self-limiting. General lesson for agent/eco sims: a stock buffer introduces hysteresis/overshoot; if you want a variable to track a driver cleanly, drive it off the flow, not an accumulating stock. Also: ground the model in known quant laws (Gibrat proportional-growth => Zipf city sizes; Krugman NEG agglomeration) and MEASURE the emergent signature (rank-size slope, Gini) rather than assert it.