It is a truth universally acknowledged that code expands to fill the available capacity to debug it. We’ll get back to software in a minute, but let’s talk about a man first.
Robert Moses was a uniquely powerful, smart, and corrupt figure in mid-20th-century New York City and New York State. He was an unbelievably prolific builder who channeled various public and private resources into the parkways, beaches, and bridges of the New York metropolitan area by every means possible. He had no problem lying, ruining people’s lives, and upending communities if it served his purposes. Under his leadership, the city gained many more roads that were wider and better. It had more bridges connecting various parts of the city. It had more parkways.
Robert Moses and his subordinates believed that all this would alleviate the traffic problems the city was already experiencing with the mass production of cars. To their surprise, as they built more roads, more cars seemed to pop up. This offset the change in capacity and left congestion unchanged. They initially believed this to be a good sign. It meant that many people were suddenly able to use the roads, and they assumed the eventual result would be a smoother flow of traffic.
They were wrong.
More roads, they learned slowly, meant more, not less, traffic. This was counterintuitive at the time, but it is well understood today that this is how traffic in heavily populated areas works. It becomes congested up to a point, and adding roads provides a very temporary solution to the problem, if that.
Academics caught wind of this phenomenon and felt the immediate need to name it. Around 1962, economist Anthony Downs formulated "Downs's Law of Peak-Hour Traffic Congestion." By the '90s, the general case became known as induced demand.
For some things, it seemed, increasing the underlying resource’s availability is never enough.
Fast forward to emerging software development patterns in light of advances in AI. We appear to be more efficient and more productive. We need fewer people to do the same job, and we do it faster than ever. We generate code. More and more code.
The more code we generate, the more bugs there are in the world. The more bugs we find, the more work these code-generating machines will have. To fix them, they will generate more code, and round and round it goes. Therefore, in this software engineer’s opinion, reports of our profession’s death have been greatly exaggerated. The industry overall will end up needing similar numbers of people. The increase in efficiency we currently feel will soon be squashed under masses of code piling ever higher and deeper. There will always be just one last bug to solve, for which the AI apologizes humbly. "It was indeed an oversight on my part," it will say. "I will fix it immediately. Is there anything else you’d like to talk about?"
