This week brought major updates and challenges in AI-powered coding tools. Developers pushed back against Anthropic’s Claude 3.7 model due to buggy code and ignored instructions, with many switching back to older versions. GitHub Copilot introduced limits for premium AI models like Anthropic’s Sonnet 3.7, requiring paid upgrades for heavy users. A new study revealed top AI models like Claude 3.5 solve fewer than half of real-world coding tasks, showing they still need human help. Meanwhile, 84% of tech leaders said AI won’t replace low-code tools, which save companies thousands of hours.

Companies like Microsoft and Google rolled out new tools for professional developers. Microsoft added GPT-4.5 and upgraded enterprise features in its Azure AI Foundry, while Google’s Gemini Robotics models showed promise for controlling robots. However, coders warned that AI assistants often create messy or incomplete solutions, especially for complex projects.