Specification Affordance

In The You On AI Encyclopedia

The command-line interface required translation into formal syntax, a transformation of the problem's structure that discarded aspects of the original intention. The graphical interface reduced but did not eliminate the translation, mapping machine operations onto visual metaphors that approximated but did not match the builder's perceptual categories. The touchscreen reduced the obstruction further. Each transition decreased the translation cost. None achieved directness in Gibson's sense, because all required translation from the perceptual language of the builder to the operational language of the machine.

The large language model removed the translation entirely. The builder describes what she perceives in the language in which she perceives it. The obstruction between the organism and the information — present in some form throughout the entire history of computing — vanished. This is the event Edo Segal describes in You On AI as the machines learning to speak our language, and the significance of the event is what the specification affordance captures.

The transformation extends beyond efficiency. It restructures perception. Prior to the specification affordance, the builder's perceptual bandwidth was consumed by terrain information — the mechanics of each implementation step, the syntactic and structural details of the code being produced. With the AI handling terrain, perceptual bandwidth is freed for destination information — the strategic trajectory, the architectural implications, the user-facing consequences of the design. The figure-ground relationship of the builder's perception inverts.

The specification affordance is not limited to software. It extends to any domain in which an AI system can translate natural-language descriptions of intent into domain-appropriate implementations: legal drafting, technical writing, analytical synthesis, creative production. In each domain, the builder's perceptual engagement shifts from the mechanics of execution to the strategy of direction. And in each domain, the same Gibsonian question arises: does the specification affordance build the perceptual sensitivities it demands, or does it presuppose sensitivities that were built under the old affordance structure and that the new structure no longer generates?

Origin

The affordance is articulated in this book's reading of Edo Segal's account of the December 2025 threshold. Segal's description of describing a problem to Claude in plain English and receiving an implementation is the empirical material; Gibson's affordance framework supplies the theoretical name.

Key Ideas

Categorically new. The specification affordance is not an improvement on implementation affordances but a different class of action possibility.

Perceptual inversion. The builder's attention shifts from terrain (mechanics of each step) to destination (strategic trajectory).

Translation abolished. For the first time, the builder describes problems in the language in which she perceives them, without translation into a formal intermediate.

Domain-general. The affordance extends to any domain where AI can translate natural-language intent into domain-appropriate implementation.

The presupposition problem. The specification affordance is most valuable to builders whose perceptual sensitivities were tuned by the implementation affordances it replaces.