In few years, the late 20th and early 21st centuries will be seen as the era in which humanity had universal computation and skills to write code, yet still relied on verbal promises and lawyer-drafted contracts—forms of language inherently vulnerable to ambiguity and reinterpretation—to coordinate high-stakes behavior.

It is important to understand why code is a critical innovation from the perspective of making contracts with each other.

First, it was the only practical way to turn an idea into an exact digital artifact. It did not merely describe a product at a high level. It specified exactly how the product would behave in practice, bugs included. For purely digital systems, code becomes the real specification.

Second, code is one of the key preconditions for enabling a new kind of credible commitment. Verbal promises and legal contracts are often ambiguous and open to reinterpretation later. Code can make commitments far more precise. That precision makes it possible to codify constraints, guardrails, fault detection, and dispute-resolution logic directly into a system’s behavior. But for that logic to matter socially, it had to be anchored to a substrate on which it could be credibly committed and enforced. That was the deeper promise of blockchains: the possibility that executable logic could become a credible means of enforcing rules rather than merely describing intentions.

Two limitations prevented this possibility of credible commitment from becoming general. First, very few people could write code well, even though human beings are exceptionally good at generating ideas. So even after the internet transformed commerce, most of society still relied on verbal agreements, archaic institutional trust, and legal instruments that remained fuzzy at the point of dispute. The power to specify systems in code remained concentrated in the hands of a small class of technical specialists (the SWEs).

Second, the commitment substrate itself remained immature. Blockchains were underdeveloped and, over time, acquired a bad reputation among many technologists because of their association with speculation, scams, and financial excess. So even where code could in principle serve as a vehicle for credible commitment, the surrounding infrastructure and legitimacy were missing.

Recent advances in AI have begun to solve the first limitation. If everyone can effectively command a personal coder, then producing code becomes cheap and abundant. At that point, writing code by hand may come to look like an aberration in human history: a brief period when people had to manually translate intent into machine-readable logic themselves. Even the practice of sharing code with each other has started to be viewed as an inefficient practice, as seen in Karpathy’s “idea file” concept.

The second limitation is still with us. Overcoming it is not just a technological challenge, but a problem of social adoption, institutional acceptance, and public perception. That will likely be a much longer fight.

Although this essay begins as a reflection on AI making code production abundant, the larger claim is more general: AI may commoditize the creation of precise digital commitments. If that happens, the real historical anomaly may not be that humans once wrote code by hand. It may be that, despite living in a computational civilization, we relied for so long on ambiguous language to coordinate the things that mattered most.