Pump Priming (Preparation)

In The You On AI Field Guide

Poincaré was insistent that the preparation could not be shortened beyond a certain threshold without altering the quality of what followed. The fifteen days of failed effort preceding the Coutances recognition were not fifteen days of waste. They were the precondition. Shortening them would have shortened the activation, and the insight that arrived on the omnibus step would not have been the same insight — perhaps not an insight at all, but a partial result requiring further work.

The distinction between passive reading and active wrestling is crucial to Poincaré's account. A mathematician could read every relevant paper, memorize every relevant theorem, and absorb every relevant technique without activating the elements in the way the creative process required. Reading is passive. Wrestling is active. The unconscious, Poincaré believed, could only work with elements that had been actively engaged, not merely passively received. This distinction is directly relevant to AI-augmented work: the builder who describes a problem to Claude and receives a solution has not wrestled; she has described. The activation, in Poincaré's specific sense, has not occurred, or has occurred at a shallower level than the struggle would have produced.

K. Anders Ericsson

The modern cognitive science of expertise — particularly K. Anders Ericsson's work on deliberate practice — provides complementary evidence for the function of effortful engagement. Expert performance is built through sustained engagement with tasks slightly beyond current ability. The effort is the mechanism that drives neural reorganization. Ericsson's framework is not identical to Poincaré's, but the overlap is substantial: both describe a form of effortful engagement that builds cognitive structures invisible to the person building them, structures that later enable perception and insight that feel effortless when they arrive.

The implication for AI-augmented work is pointed. If the depth of creative insight is correlated with the depth of prior struggle, and AI tools compress or eliminate the struggle, then the resulting insights may be systematically shallower than those produced through the unaugmented process. This is the argument the Poincaré simulation volume develops across its middle chapters: that eliminating the struggle is not an efficiency gain but the removal of the cognitive process that produces the most valuable class of creative results.

Origin

Poincaré used the pump metaphor explicitly in Science and Method (1908) to describe the paradoxical productivity of apparently wasted effort. The metaphor's mechanical precision was characteristic: Poincaré, trained as a mining engineer before becoming a mathematician, was comfortable with physical analogies that captured functional relationships without requiring metaphysical commitments about consciousness.

Key Ideas

Failure is the work. Every unsuccessful approach activates an element, establishes an association, eliminates a region of the combinatorial space. The visible output is zero; the invisible output is the loading of the unconscious with material for incubation.

Activation requires engagement, not description. Reading about a problem does not prime the pump. Wrestling with it does. The distinction determines whether the unconscious has material to work with during the incubation phase.

Depth of preparation correlates with quality of illumination. Shallow preparation produces minor insights. Intense, sustained engagement produces the specific quality of recognition that restructures understanding.

The priming cannot be purchased or outsourced. The AI can produce output. It cannot build the cognitive architecture within which the builder's own subsequent insights can form, because that architecture is the product of the builder's own engagement.