AI's Impact on Coding: Implementation vs. Organizational Problem Solving

The Shifting Definition of "Coding Solved"

The idea that "coding is solved" often refers to the increasing ability of language models to translate well-defined problems into runnable code. This capability significantly reduces the time required for tasks like building REST APIs or implementing React components, making the output generation process much faster.

Demo Apps vs. Enterprise Challenges

Building demo applications with clear requirements and few constraints highlights AI's effectiveness. However, in a mature organizational setting, adding a simple feature involves answering numerous non-coding questions related to service ownership, existing APIs, security, architectural patterns, team responsibilities, downstream system impacts, and business constraints. These questions require deep organizational understanding, which AI currently lacks.

Mistaking Bottleneck for Discipline

Historically, writing code was a major bottleneck in software development, leading to the equation of software engineering with implementation. AI's advancements are making implementation dramatically cheaper, thereby moving the bottleneck to higher-level concerns. This shift redefines what becomes scarce and where attention is focused within the development process.

Layers of Software Engineering

Software engineering encompasses several layers: defining business objectives, product design, solution design, and implementation. While current AI models are becoming highly proficient at the implementation layer, the other layers, which involve understanding complex problems and organizational context, remain largely untouched by AI's current capabilities.