“氛围编码”2年攒下的烂摊子，正在逼我重新手写代码

Core Insights - The emergence of AI coding tools has sparked debates about whether machines can replace human developers, with some praising their efficiency while others caution against potential limitations in code quality and system stability [1] - A developer named mo shared his experience of relying on AI for "vibe coding" over two years, ultimately realizing that while AI-generated code may seem reasonable in parts, it struggles with overall structure and long-term maintainability, leading him to revert to hand-coding [10][11] Group 1: Developer Experiences - Many developers follow a similar journey with AI coding tools, initially impressed by their ability to handle simple tasks and later complex ones [2][6] - As developers assign more complex tasks to AI, they begin to notice flaws, leading to frustration and a tendency to blame themselves for the AI's shortcomings [7][9] - The reliance on AI for coding can result in a lack of understanding of the overall system architecture, as AI-generated code often fails to consider the broader context [10][12] Group 2: Educational Concerns - There is growing concern among educators that AI's ability to perform simple tasks too well may lead novice programmers to skip essential foundational training, hindering their long-term development [11][12] - Teachers emphasize the importance of students writing their own code to build understanding and intuition, as relying on AI can prevent the internalization of critical skills [12][13] - Experienced engineers note that while AI can boost productivity in the short term, it may leave developers stuck at a lower skill level without the necessary growth to reach their full potential [14][15] Group 3: Caution in AI Usage - Developers are increasingly adopting a more cautious approach to using AI, recognizing that while it can save time, it often introduces technical debt that accumulates over time [14][15] - The reliance on AI can weaken a developer's understanding of code, leading to difficulties in troubleshooting and a lack of a mental model of the codebase [16][17] - As developers become more dependent on AI, their productivity may plummet when they are unable to access these tools, highlighting the risks of over-reliance [16][17]