A software developer who builds tools for musicians is wrestling with a question that has become common across the industry. As large language models grow more capable, what is left that only a human can do? The answer, for one developer at least, has less to do with technical skill than with the spaces between the things that machines can already replicate.
The Moat Keeps Moving
Calvin Flegal, a developer who has been thinking about the future of his career in software, describes a pattern that many in the industry now recognize. At first, it seemed as though large language models were not great programmers. That phase was brief. Then the argument shifted to product vision, the idea that a human could still conceive of something a machine could not imagine. But even that started to erode once it became possible to feed product analytics into a model and receive a refined vision in return.
The pattern continues. Each defensible advantage that a software developer once claimed has a way of being absorbed by the models. The problem is not unique to Flegal or to software. He notes that it is far from it.
What Remains Is Human
What Flegal is left with are core beliefs about humans and about the things people value in the work they experience. He turns to music for the metaphor. There are well-known sayings in that world: drum machines have no soul. It is not the notes but the space between the notes. The notes do not teach you how to make the music.
People love music for its imperfections. They love it because it brings them together. Flegal acknowledges that he may soon be fooled by what an LLM produces, and he suspects the time is coming when he will be truly moved by a piece of music only to learn it was entirely made by a machine. When that moment comes, he expects to be upset, not because the output was poor, but because the thing he was after was the humanity of it all.
A Small App and a Big Question
Flegal has been building a project called Music Mini Games, an iOS application designed to help musicians improve their skills. A Mac version is awaiting review. He has used Claude to replicate much of the functionality in a web version, though he admits the web version may be of low quality and he has not spent much time with it.
The app is his vision, but he recognizes it still belongs to the old world. He raises a pointed question about what happens to something like it in an era where a person could have their education instantly conjured, changed on the fly, and personalized based on real-time feedback about how they are doing. A fixed curriculum, even one with personalization, faces an uncertain future when a model can adapt to a learner's every move.
He references Nikita Bier, a product builder whose philosophy he admires, and wonders whether his little project should try to become what Bier describes or simply accept that the value of a small, curated tool may decline quickly.
The Space Between the Notes
Flegal's conclusion is not pessimistic, but it is quiet. He hopes there will continue to be space for the imperfect work that software makers produce. He hopes his own projects will remain imperfect. He hopes that some people will still value the space between the notes, the thing that machines have not yet replicated because it is not a product feature but a human choice.
The essay is a personal meditation, not a technical analysis, and it reads as such. But the question it raises is becoming harder to ignore. If a model can write code, conceive products, adapt curricula, and produce music that moves people, what is genuinely human left in the work that developers and creators do? Flegal does not offer a confident answer. He offers a hope that people will still want to hear the tunes that someone imperfect made on purpose.