One of Gould's most persistent arguments challenged the widespread assumption that evolution is progressive—that life has a directional tendency toward greater complexity, intelligence, or any other seemingly 'advanced' feature. The Full House analysis demonstrated that apparent progress is usually a statistical artifact: life began simple (left wall—minimum complexity for self-replication), variance expanded in all directions, leftward expansion blocked by wall, distribution appeared to move rightward. But the mean of life's complexity barely moved across three billion years—bacteria remain the most abundant, ecologically dominant, and metabolically diverse organisms. The right tail extended (multicellular life, nervous systems, consciousness) but the tail is not the trend. The trend is expansion, not direction. Applied to AI, the framework challenges the assumption that intelligence is a substance accumulating in the universe, flowing directionally from atoms to algorithms to AGI. Instead: information-processing capability expands from constrained origins (simple chemistry, basic computation), variance increases as new forms emerge, some forms happen to occupy high-complexity positions, but the mean may not move and the right-tail occupants are not destinations the process aimed toward. The river of intelligence flows, but the river does not determine where it flows—the channel is carved by contingent landscape features.
The progressive view of evolution—the ladder climbing from simple to complex, from lower to higher—is not a scientific hypothesis but a cultural inheritance from the Great Chain of Being, Christianized teleology, and Victorian faith in progress. Darwin himself resisted progressivism, emphasizing that natural selection has no direction beyond local adaptation to immediate conditions. But popular Darwinism absorbed progressivism from the culture that received it, and the ladder imagery (Haeckel's tree, the March of Progress illustration) became evolution's public face despite contradicting its actual mechanism.
Gould's distributional analysis—examining variance over time rather than right-tail extremes—revealed the pattern consistently. Across the history of life, across the history of baseball, across any system beginning from a constrained starting point, the appearance of directional progress is produced by variance expansion from a wall. The right tail extends not because the process aims rightward but because the wall prevents leftward expansion. This is not progress—it is asymmetric variance expansion, and confusing the two is the fundamental error of progressivist thinking.
Applied to Edo Segal's river of intelligence metaphor, the non-progressive framework accepts that something flows (information processing capacity, organized complexity, anti-entropic structures) while contesting that the flow has a predetermined destination. The river follows gravitational gradient (thermodynamic tendency toward dissipative structures far from equilibrium) but the specific channel depends on geological contingencies. The tendency toward complexity is real as statistical phenomenon. The specific realizations (biological neurons, transformer architectures, whatever comes next) are contingent products of specific landscapes.
The practical consequence for the AI moment: if complexity is not progressive—if the transformer architecture and large language models are not inevitable expressions of intelligence accumulating toward AGI—then the specific future depends entirely on specific choices made now. The dams builders construct are not refinements of inevitable trajectories but constitutive interventions determining which of many possible futures actually materializes. The bush branches. The specific branches surviving are determined by the landscape, and the landscape is shaped by contingent choices, and the choices are being made by specific humans at this specific bend in the river.
Gould's anti-progressivism was a lifelong commitment, articulated across dozens of essays and multiple books but crystallized most clearly in Full House (1996). The position was controversial—many biologists accepted directionality as empirical fact (brain size increasing over mammalian history, for instance). Gould's response: local trends are real, but there is no global trend toward complexity. The appearance of global progress is produced by variance expansion from constrained origins. His final major work, The Structure of Evolutionary Theory (2002), defended the non-progressive view as essential to Darwinism properly understood.
Life began at the left wall. Simplest possible self-replicating chemistry defined the starting constraint; all subsequent variance is expansion from that origin.
Bacteria are the mode, not the exception. The most abundant and ecologically successful organisms are the simplest—complexity is a small right-tail phenomenon.
Apparent progress is variance expansion. When a distribution expands from a wall, the right tail extends without the mean necessarily moving—statistical artifact of constraint.
No global trend toward intelligence. The river does not flow toward consciousness; consciousness is a rare contingent outcome occupying one branch of the expanding bush.
Specific outcomes are contingent, not necessary. The transformer, LLMs, AGI-trajectories are not inevitable expressions of accumulating intelligence but contingent products of specific choices in specific landscapes.