Saying Goodbye to Agile

Agile has dominated the industry, but in the end it was never quite clear what it actually is. Whenever failure occurred, it was always explained away with "that wasn't real Agile," and the manifesto reads more like a set of vague maxims than a concrete process.

The core ideas it champions—iterative development and customer involvement—had already existed in research and practice since the 1970s. The limitations of waterfall were also nothing new; early literature such as Royce (1970) and Bell-Thayer (1976) had already clearly pointed out the importance of prototypes, continuous customer involvement, and specifications and design.

What's becoming important again now is spec-driven development. Because LLMs are vulnerable to ambiguous input, clearer specifications and documentation determine the quality of code generation. The author argues that, in contrast to Agile's slogan "working software over comprehensive documentation," we now need to return to the view that comprehensive documentation produces working software.

The conclusion is clear.

• Rather than worshipping Agile as an absolute principle, we should look at actual engineering principles and historical context.
• Iteration and feedback are nothing new, and documentation and specifications remain essential.
• In the AI era, clear specs matter more than vague processes.