Member-only story
Intent-Oriented Programming: Bridging Human Thought and AI Machine Execution
A potential solution to looming Vibecoding challenges.
I’ve been a vocal proponent of “vibecoding” for the past year — the approach where developers express their intentions to AI systems that then generate functional code. While Vibecoding offers tantalizing glimpses of a more intuitive development process, it comes with significant challenges: inconsistent outputs, difficulty maintaining larger systems, and the “black box” nature of generated code.
What if there was some middle ground? A structured way to express intent that maintains human understanding while guiding AI generation more effectively?
Last week, I stared at a seemingly simple problem that had ballooned into hundreds of lines of code across multiple languages. What started as a straightforward user authentication flow had become an unwieldy tangle of implementation details. I couldn’t help but think: “There must be a better way to express what I want this system to do.”
This frustration led me down a path of exploring an idea that’s been quietly developing in various corners of software engineering: intent-oriented programming. It’s an approach that prioritizes describing what a system should accomplish rather than how it should be implemented.
