Developers say AI coding tools work—and that's precisely what worries them

Software developers have watched AI coding tools evolve from advanced autocomplete into agents that can build applications from text prompts. Ars Technica spoke to several professionals who largely agree the technology works, but their reactions mix enthusiasm with unease.

Some developers describe substantial productivity gains: agents that write code, run tests, and help debug can speed prototyping and modernize legacy systems. Others warn of risks such as hallucinations, accumulating technical debt, and the phenomenon called 'vibe coding' coined by a former researcher, where developers accept AI-generated code without fully understanding it.

Adoption also varies by context: large enterprises face cost and legal hurdles to deploy more capable models, while many workers get limited features embedded in existing products.

ai coding tools, advanced autocomplete, agents that can build applications, text prompts, productivity gains, agents that write code, run tests, help debug, speed prototyping, modernize legacy systems, hallucinations, accumulating technical debt, vibe coding, cost and legal hurdles