The way that organizations and organisms anticipate the future is by taking signals from the past, most the time.

Kelly’s line frames “anticipation” as a fundamentally backward-looking process: whether in biology (nervous systems, immune responses, evolution) or in institutions (markets, bureaucracies, cultures), forecasts are typically built by detecting patterns that have already occurred and treating them as predictive signals. The quote also hints at a limitation: reliance on past data works “most of the time,” but can fail when conditions change, novelty appears, or rare events dominate. In Kelly’s broader techno-social thinking, this supports a pragmatic view of prediction—useful, adaptive, and probabilistic rather than certain—and suggests that resilience comes from updating models quickly when the past stops being a good guide.