The reversal goes deeper than skill obsolescence. A builder who has used AI for years has not simply gained new capabilities — she has lost old ones. The capacity for sustained debugging, the tolerance for friction, the ability to sit with a problem for hours without reaching for a tool — all have weakened through disuse. When the tool is removed (power failure, subscription lapse, medium evolution in a direction that no longer serves her), she discovers the empowerment was conditional. It depended on the continued availability of the extension.
The reversal operates at multiple scales. For the individual builder, it is the atrophy of unassisted capability. For the organization, it is the loss of institutional memory about how work was done before AI — how problems were scoped, how code was reviewed, how quality was evaluated. For the profession, it is the disappearance of the pathway through which expertise was traditionally built. The apprentice who learned by wrestling with problems AI now solves does not become a journeyman; she becomes a supervisor of machine output whose expertise cannot deepen through experience she no longer has.
The reversal is not predicted by the extension narrative. It requires the tetrad's simultaneous view: empowerment and dependency operating in the same medium at the same moment, visible only when both are held in mind together. You On AI documents the enhancement extensively; the reversal appears only in oblique confessions — the reluctance to attempt anything without Claude, the slight panic when connectivity drops, the shift in confidence about unassisted work.
The structural question is whether the reversal can be interrupted by deliberate practice — by building dams that preserve unassisted capacity against the medium's amputating force. The answer is partial. Individual practice can slow the reversal for the practitioner who maintains it. It cannot reverse the social-scale consequences, because those depend on institutional structures that must be built at scale. The reversal is not inevitable. It is likely, unless structures are built against it.
An application of McLuhan's tetrad to AI, developed explicitly in the present analysis. The general structure — technology produces its opposite at the extreme — is articulated throughout Laws of Media (1988). The specific application to cognitive extension represents an extrapolation from McLuhan's analyses of mobility technology, communication technology, and entertainment media to the novel case of thought-extension.
Structural, not contingent. Reversal is produced by extension itself at the extreme — not by misuse or bad design.
Deeper than skill loss. The builder loses capacities she once had, not merely fails to develop new ones.
Multiple scales. Individual atrophy, organizational memory loss, professional pathway disappearance — all operating simultaneously.
Invisible from within. The empowerment is celebrated while the dependency accumulates — the tetrad's simultaneous view reveals what sequential analysis misses.
Interruptible, not preventable. Deliberate practice can slow individual reversal; institutional structures must address the social scale.