AI coding is now everywhere but not everyone is convinced (www.technologyreview.com)

The rise of AI-powered coding tools, heavily backed by tech giants like Microsoft and Google, has sparked a fervent debate within the software development community. While executives tout productivity boosts from AI-generated code—reporting claims that a significant portion of their code is now automated—many developers are expressing skepticism. A series of studies and developer testimonials reveal a troubling trend: the productivity gains promised by these tools may be overstated, with some developers experiencing slower work rates and increasing frustration with the software’s limitations. For instance, a notable study found that while developers felt faster using AI, objective tests indicated they were actually less efficient. This disillusionment stems from several key issues. AI tools excel in producing boilerplate code and fixing bugs but struggle with larger, complex tasks due to limitations in their context window, often leading to inconsistent or poorly designed code. As developers increasingly rely on AI, there is a growing concern about accumulating technical debt, where shortcuts taken for speed result in harder-to-maintain software. This fosters a cycle of complexity that could hinder long-term project sustainability, as the tools generate “code smells”—subtle flaws that are challenging to identify—over more evident bugs. As such, the relationship between AI coding tools and developer productivity remains murky, reflecting a landscape where expectations may not align with reality.