Brian 'Beej' Jorgensen Hall, a CS instructor at Oregon State University-Cascades and longtime industry veteran, has published a philosophical essay that cuts to a question dividing the developer community: when an AI writes your code, did you actually make anything? Hall — known for his widely-read programming guides — opens with an elaborate misdirection, presenting a sci-fi novel excerpt, a woodcut painting listed at $1,300, and a Rust roguelike as examples of his own prolific creativity, before admitting all three were AI-generated. Only his rebuilt front deck was genuinely his own work — and even that was built by paid craftsmen.
The essay draws a sharp line between tools that extend human agency and AI systems that replace it. A compiler or a hammer amplifies what you personally create; asking Claude Code to generate a codebase is closer to hiring a contractor — something Hall would never describe as his own work. 'As a manager of LLMs: my agents built this,' he writes, 'and that, for me, has very little weight in terms of making.' He acknowledges the view isn't universal — plenty of people say 'I built a deck' when contractors did the labor — but Hall finds that framing personally dishonest.
The practical consequences run further than philosophy. He won't MIT-license AI-generated code he initiated but didn't write, opting instead for the Unlicense: he can't honestly grant rights over work that isn't his. He also recounts building a flash card app for his wife entirely by hand — using Claude only to look up API documentation, explicitly refusing to let it write code — in a process he estimates took 50 times longer than delegation would have. He ended up with 177 lines and, by his account, more pride than anything Claude has produced for him. Pride, in his framing, requires having done the work yourself.
That last point is worth sitting with for anyone building on top of agentic coding tools. The pitch for Claude Code, Devin, and their successors is partly that developers can claim ownership of what emerges — that directing an agent is itself a form of authorship. Hall's essay is a precise challenge to that premise: if the doing belongs to the agent, so does the credit. He isn't telling anyone else to stop using these tools, and his conclusion is explicitly personal. But a working developer who reaches for Claude Code regularly and still can't bring himself to claim the output as his own is a useful corrective to the industry's habit of papering over the question.