We have become strangely preoccupied with what AI can produce: code, images, essays, and even entire products.
But production was never the whole story.
Creative and technical work have always happened across layers: defining the problem, choosing a direction, producing the artifact, reviewing what comes back, and bringing the parts into a coherent whole. We tend to recognize expertise in the finished thing because that is the part we can see.
AI is making the rest harder to ignore.
The person using AI is not removed from the work. They move to a different layer of it.
Building my portfolio with Cursor made that shift unusually visible. I spent less time implementing every detail myself and more time breaking problems apart, defining requirements, deciding what to delegate, reviewing solutions, rejecting weak ones, and protecting the overall direction.
Debugging did not disappear either. Cursor could trace code, identify likely causes, and propose fixes, but I still had to decide what was actually wrong, whether the fix made sense, and what else it might break.
The artifact became easier. The judgment became harder.
Writing makes this obvious. People often say they hate AI-written content. Sometimes the problem is not that AI wrote it. It is that it did not need to be said.
Writing is no longer the hard part. Having something worth writing about is.
Software is not so different. Generating code is only one part of building a product. Someone still has to decide what should exist, how to divide the problem, which constraints matter, and when something that technically works is still wrong for the larger system.
AI can produce something polished and completely wrong surprisingly quickly.
Everyone becomes more responsible for defining the problem well.
Cursor generated a significant amount of code for my portfolio. But saying Cursor built it would miss most of the work.
I made hundreds of design decisions: what the site should feel like, which interactions belonged, which ideas had to be abandoned, what deserved another pass, and what needed to remain untouched. I divided the work, tested the results, debugged what failed, and kept bringing local changes back to the larger vision.
Cursor accelerated production. My work became orchestration.
AI may be accelerating this shift, but the pattern is familiar. Every major technological change has altered where humans create value. Industrial machinery reorganized physical labour. Computers shifted work from calculation toward analysis. Spreadsheets did not eliminate accountants; they shifted more of the work toward interpreting results, modelling outcomes, and advising decisions.
AI may be doing something similar. Not by making people less important, but by making the deeper parts of the work more visible: judgment, direction, taste, coordination, and the ability to recognize what is worth making.
As AI becomes better at producing artifacts, will our value increasingly lie in deciding which artifacts are worth producing in the first place?