In 1954, IBM predicted that FORTRAN would virtually eliminate coding and debugging.
Seventy-two years later, we're having almost exactly the same conversation about AI.
And the interesting part isn't that the prediction was wrong. In some ways, it was right.
We've been here before
FORTRAN eliminated enormous amounts of manual work. Programmers stopped translating mathematical formulas into machine instructions.
But programming didn't disappear. It moved up a level.
The same pattern repeated with higher-level languages, libraries, frameworks, and development tools. Every generation automated work that previously required human effort.
Now AI agents are doing something similar, only much faster.
But there's something different this time
The evidence is more complicated than either the AI enthusiasts or skeptics would have you believe.
In a 2025 randomized study by METR, experienced developers working on familiar open-source repositories actually took 19% longer when using AI tools.
The fascinating part? Those same developers believed AI had made them faster.
That doesn't mean AI makes developers slower in general. The study was limited, the tools have improved, and the tasks mattered enormously.
But it illustrates something important: feeling more productive and being more productive aren't necessarily the same thing.
Meanwhile, the length of tasks AI agents can complete autonomously has been increasing rapidly.
So perhaps the real question isn't whether AI will replace programming.
It's what programming becomes when writing code is no longer the expensive part.
A developer's Tuesday in 2031
I started imagining an ordinary working day five years from now.
Your coding agents have been working overnight. Three have completed their tasks. A fourth stopped because it encountered a question that required understanding the business rather than generating code.
You spend your morning reviewing work, resolving ambiguities, and deciding what should happen next.
You barely write any code.
Are you still a programmer?
I think so. But your job looks very different.
I explored this idea in considerably more detail, looking at historical predictions, developer surveys, research into AI productivity, and five predictions for 2031.
Read the full illustrated article on hrih.si →
I also publish experiments, technical deep dives, and independent projects at hrih.si, where you can subscribe to Letters from the workshop.
Top comments (0)