Programming has changed a lot through the years. Over the last few decades it seems like programmers have been competing over who uses the most arcane and inscrutable technology and process - see xkcd Real Programmers for reference. These days there’s a cultural shift to prove either we can be “orchestrators” or “supervisors” of generative Artificial Intelligence agents or we’re unnecessary and solo founders can “vibe code” their entire business. These extremes are predicated on the assumption that a computer program is a collection of lines of code.
I’m a software engineer working in industrial automation and in this article I will argue that a program or digital system is not its source code or binaries, but the constellation of client and user requirements shaped by business and technical constraints, as it exists in the minds of the people responsible for it. I’ll support this idea with key bibliography as well as by recounting a few personal experiences in software project onboarding.

The author experimented with building a time zone conversion app using OpenAI’s Codex, a coding agent. The experience was positive, with Codex successfully implementing the requested features, including offline functionality and a responsive design. However, the author still had to review and refine the code, highlighting the importance of human oversight in AI-assisted development.