Glyph Lefkowitz argues that programmers should take creative workers’ objections to generative AI seriously: software is creative work too, even when the task is ordinary and commercial. A commissioned portrait, advertising copy, and a custom WordPress theme do not stop being creative because they serve a practical purpose.
His strongest argument concerns how people acquire skill:
- Mundane work is an apprenticeship. Repeated practice and attention to a problem give people the understanding needed to improve their craft.
- Frustration can produce better designs. Lefkowitz describes Twisted’s
Deferred—a way to organize asynchronous operations—as an aesthetic response to the tedium of manually passing callbacks, the functions that run when an operation finishes. Experiencing that awkwardness helped him see a better approach. - Automation and loss of understanding are different things. He explicitly allows abstraction and automation. His objection is to using AI to eliminate the understanding and creative decisions involved, rather than elevate them.
The consequential question is not whether every piece of code deserves a gallery wall. It is what happens to the development of taste and judgment when organizations remove the everyday work through which people acquire them. Eliminating entry-level tasks while demanding experienced developers risks consuming an apprenticeship system without replacing it.
The essay makes a broader case for resisting AI in programming, but calling programming art does not by itself establish that AI assistance is wrong. Reviewing, debugging, and experimentation can also develop judgment. The useful test is whether a workflow strengthens those activities or turns them into approval of output nobody meaningfully understands: defend the learning and agency embedded in mundane work, not tedium for its own sake.