Frank
In 1952, a Navy programmer built a tool that could write its own instructions, and the engineers around her spent three years refusing to use it because they were convinced a machine couldn't possibly understand math.
The Story
Grace Hopper gets the headline credit, and she deserves it. But the part of the story that gets skipped is the three years nobody would touch her compiler. The A-0 System, finished in 1952, could take symbolic code and translate it into machine instructions automatically. That was the whole point. Write it once, run it anywhere. She handed it to programmers and they handed it right back.
The objection wasn't technical. It was emotional. Programmers in 1952 believed, genuinely and deeply, that a machine could not write code as efficiently as a human who understood the hardware. They weren't dumb. They were just wrong. They had spent careers learning to think in binary, optimizing every instruction by hand, and they weren't about to let a piece of software do it faster. Their skill was the product. The compiler threatened to commoditize it.
So they sat on it. The U.S. Navy had a working compiler and the people who could use it refused to. For three years. Hopper kept pushing anyway, kept showing the output, kept documenting how the compiler's code matched or beat hand-written output. Eventually COBOL came out of that same philosophy in 1959, and the whole industry had to admit she was right. By then she'd already moved on to the next thing they wouldn't believe.
The Hidden Principle
The resistance wasn't about capability. It was about identity. The programmers who rejected the compiler weren't protecting quality, they were protecting status. Hand-coding in assembly wasn't just a skill, it was proof you understood something most people didn't. A tool that did it automatically made the proof worthless. That's a completely different problem than "does this work."
This shows up every single time a tool threatens to automate something people have built their self-worth around. The fight looks technical. It never is.
What This Means Today
Right now, the same argument is happening word for word about AI and coding. Senior engineers calling GitHub Copilot a crutch, saying LLM-generated code is sloppy, saying you can't trust what you didn't write yourself. Some of that is true. A lot of it is the same 1952 energy, dressed up in code review comments. The compiler was also "sloppy" compared to a careful human. Didn't matter. Scale won.
The principle is this: every generation of technical people has one tool they refuse to adopt because adopting it means admitting the thing they mastered can be abstracted away. They're always eventually wrong. The people who figure that out early don't win because they're smarter. They win because they stopped protecting yesterday's skill set and started building on top of the new one.
The tool you're refusing to learn is probably the one writing your replacement.
