January 24th, 2025

AI Coding Is Based on a Faulty Premise

The article warns that increasing reliance on AI in software development may lead to poor quality, echoing past software crises. It emphasizes the necessity of human intuition and communication in coding.

Read original articleLink Icon
AI Coding Is Based on a Faulty Premise

The article discusses the increasing reliance on AI in software development, particularly the trend of replacing human programmers with AI tools. The author expresses concern that this approach overlooks historical lessons from the software crises of the 1990s, where misunderstandings and miscommunications led to project failures. The piece highlights the inadequacies of the waterfall model of software development, which often resulted in significant discrepancies between user needs and delivered products. It emphasizes that successful software development relies on human intuition, communication, and iterative feedback, as outlined in the Agile Manifesto. The author warns that using AI as a primary coding tool could lead to similar pitfalls, as non-programmers may not fully understand the code generated by AI, potentially resulting in poor software quality. The article concludes by advocating for the continued importance of skilled human developers in the software creation process, suggesting that removing them could lead to a regression in software quality.

- The reliance on AI in coding may lead to a regression in software quality.

- Historical lessons from the 1990s software crises highlight the importance of human intuition and communication.

- The waterfall model of software development often resulted in project failures due to miscommunication.

- The Agile Manifesto emphasizes iterative feedback and human involvement in software development.

- Replacing human developers with AI could result in misunderstandings and poor software outcomes.

Link Icon 2 comments
By @trescenzi - 3 months
A friend of mine dropped out of a CS program because he couldn’t quite get programming languages. He’s always said he really wished he could just program with English. And now here we are. We can program with English. But, as the article says, the challenge has never been writing the code[1]. The hard part is actually accurately describing what you’re trying to build. And the ironic thing is that English actually makes that harder as it is significantly less precise. As the article says that’s not to ignore the power of generative AI in software development. There will be faster iterations, and more code written. It just still doesn’t solve the hard part, and in some situations it might even make the hard part harder.

[1] Yes of course there are some cases where the code is the hard part. The majority of software however that is not the case.