Developers say AI coding tools work—and that’s precisely what worries them (arstechnica.com)

Recent advancements in AI coding tools have transitioned them from basic autocomplete functions to capable systems that can build entire applications from simple text prompts. Notable tools like Anthropic's Claude Code and OpenAI's Codex have demonstrated capabilities to write code, run tests, and even debug with minimal human oversight. OpenAI has taken a significant step by using Codex to improve itself, highlighting the tool's advanced functionalities. This evolution raises questions within the AI/ML community about the balance between capability and the hype surrounding these technologies. While many developers acknowledge the effectiveness of these AI tools, there is a palpable concern about the marketing claims associated with them. Some professionals express skepticism, noting that while tools like Codex are revolutionary, they may not meet the lofty expectations set by industry narratives. Others, however, see a dramatic "step-change" in AI capabilities over recent months, evidenced by a reported tenfold increase in efficiency for complex software tasks. The discourse reflects a dichotomy within the developer community, where excitement about the potential of AI is tempered by caution regarding the overarching claims in an evolving landscape.