Is AI fundamentally changing how value gets produced?

The commodity age — the long stretch of industrial modernity Marx analyzed — was organized around objects: things produced at scale, identical in form, possessable, stockpileable, exchangeable through their material substance. Production was material, distribution was logistical, value lived in the object. Even when knowledge entered this economy (textbooks, encyclopedias, recorded music) it was knowledge-as-object: stamped, identical, stocked.

The token age does not abolish the commodity age but layers a different productive logic over it. Production is contextual: each output is generated for the immediate use, not stamped to a template. Distribution is non-logistical: the token does not need to be moved because it is generated where it is consumed. Value lives in the relation between token and receiver, not in the token itself. The form of the productive thing is linguistic rather than material — strings of text, audio, image, code, generated at point of use.

This periodization helps explain why intuitions formed in the commodity age misfire. Commodification predictions (price collapse, standardization, deskilling) describe what happens when previously bespoke objects become mass-produced. Tokenization produces different effects: not price collapse but inflationary devaluation of the token-class as a whole, not standardization but extreme contextual variation that simulates customization, not deskilling but a transformation of skill from production to validation. Does AI abundance actually devalue knowledge itself? is a token-age effect that has no clean commodity-age analog.

The strongest counterargument: this is just digital production, not a new age. But digital production prior to AI still produced storeable artifacts (files, programs, recordings). Tokens generated on demand are not storeable in the same sense — even when captured, the captured artifact is a snapshot of a process, not a productive object. The category is genuinely new.