Daily Productive Sharing 1251 - Going Back to Using My Brain

One helpful tip per day:)

After two years of using large language models (LLMs) to assist with coding, Alberto Fortin realized they were not as efficient as expected. He decided to shift his approach—taking back control of the development process and using LLMs solely as tools:

1. Cursor proved to be of limited help in real-world programming. When he pasted in error messages, the AI would suggest fixes, but often ended up causing new issues elsewhere.
2. The more detailed and specific the problem, the harder it was for LLMs to provide genuinely useful solutions.
3. He wasn’t blindly relying on AI to write code, but his focus on speed led him to neglect thorough reviews, resulting in a “just keep building” mentality that sacrificed quality.
4. His codebase had become messy: two nearly identical service files in the same directory, with inconsistent method names and parameters.
5. He used Gemini mainly for its larger context window, which allowed it to reference more existing code. But even that wasn’t enough to solve deeper structural issues.

If you enjoy today's sharing, why not subscribe

Need a superb CV, please try our CV Consultation

使用 LLM 工具辅助写代码两年之后，Alberto Fortin 发现这些工具并不如预期的高效，他决定回退一步，由自己主导写代码，而仅仅让这些工具辅助自己工作：

1. Cursor 现在帮不上太多忙。他粘贴报错信息后，AI 给出修复建议，但别的地方又出错了。
2. 问题越细节，LLM（大语言模型）越难给出真正有效的解决方案。
3. 他并不是盲目地让 AI 写代码不看结果，只是为了追求速度，没真正坐下来认真审查代码。一直在“造造造”。
4. 同一个目录下有两个服务文件，名字相近，显然做着类似的事，但方法名不同，参数也不一致。
5. 他主要用 Gemini 是因为它上下文窗口更大。每次要写类似的文件时都明确让 AI 参考现有例子。但这还不够。

如果你喜欢的话，不妨直接订阅这份电子报 ⬇️