The 70-Year Cycle: Why AI Coding is an Evolution, Not a Disruption

The current surge in AI-assisted coding is often framed as a sudden disruption that threatens to replace human developers. However, a retrospective analysis published by 36Kr suggests this is not a new phenomenon, but rather the latest chapter in a 70-year effort to automate the act of programming.

Drawing on historical patterns, the analysis argues that the industry moves through a cycle of "rise and fall" paradigms. While modern Large Language Models (LLMs) have brought pair programming to the masses, they follow a long line of ambitious attempts to decouple human intent from the manual labor of writing syntax. According to 36Kr, every time pioneers believed a specific approach would finally succeed in automating the process, "reality offered far more complex answers."

A History of Automation Attempts

The quest to automate code began long before the era of neural networks. In the 1980s, MIT developed the "Programmer's Apprentice" project. This initiative served as an early precursor to modern AI pair programming by attempting to duplicate the actual problem-solving skills humans use when coding, rather than simply generating text.

By the 1990s, the focus shifted toward the representation of logic. Microsoft introduced "Intentional Programming," a project led by Charles Simonyi. This approach attempted to eliminate traditional textual source code files entirely. Instead, it aimed to capture high-level computational intent through symbolic abstractions, such as Abstract Syntax Trees (ASTs), allowing developers to manipulate the logic of a program without wrestling with the constraints of a text editor.

The Complexity Gap

These historical efforts provide critical context for the current discourse surrounding tools like GitHub Copilot. The recurring theme across seven decades is the tension between the promise of total automation and the inherent complexity of software. Each era's tools promised to liberate the programmer, yet each eventually met the reality that software is not merely about generating lines of code, but about managing intricate logic and evolving requirements.

This perspective suggests that the narrative of the "replacement" of human developers is a recurring industry trope rather than an imminent certainty. The challenge has consistently remained the same: the gap between a human's high-level intent and the machine's execution.

What Remains to be Solved

As the industry integrates LLMs into the development workflow, the primary question is whether these tools represent a fundamental break from the past or simply a more efficient version of previous paradigms. While current AI can generate boilerplate and suggest functions with unprecedented speed, the core struggle of software construction—managing complex system architecture and ensuring logical correctness—remains a human-centric task.

Observers should watch whether current AI tools can move beyond pattern recognition to solve the deeper problem of "intent" that plagued projects like Intentional Programming. Until AI can independently manage the complex logic that humans provide, the developer's role will likely evolve rather than disappear.