Why the push for Agentic when models can barely follow a simple instruction?

The article discusses the challenges of implementing Agentic coding models when current models struggle with basic instructions. The conversation on a tech forum highlights concerns about the reliability of models like GPT5 and Gemini Pro in following simple coding instructions accurately. Users express skepticism about the hype around Agentic coding, questioning the practicality of trusting models for complex tasks given their current limitations. Suggestions include using Markdown files for persistent references and guiding AI agents through tasks to ensure accuracy and consistency in coding tasks. The discussion emphasizes the importance of managing and guiding AI models rather than relying solely on their autonomy.