AI's Hidden Gift: We're Rediscovering the Practice of Coding
Source: Dev.to
Introduction
Developers are searching for the best way to get working code from large language models (LLMs). To avoid falling into the “vibecoding” trap, many are adopting more structured approaches.
Spec Kit and the Four Guiding Questions
A recent conversation highlighted Spec Kit and its Specify → Plan → Tasks → Implement process, which aligns with GitHub’s approach to coding with AI. The article The AI development trap that wastes your time (Samuel‑Zacharie FAURE, Oct 30 ‘25) presents four questions to help regain control after too much prompting:
- Do I understand exactly the specifications I’m trying to implement, or the bug I’m trying to solve?
- Do I have an exact plan for implementing my changes?
- What is the current abstraction level to which I should be prompting now?
- Which other information am I lacking?
These questions capture the essence of disciplined coding.
Returning to Core Coding Practices
To obtain better results from AI and other tools, we need to return to the fundamentals of coding:
- Understand the problem to solve.
- Decompose that problem into smaller, manageable pieces.
- Think at the appropriate level of abstraction.
- Ask enough clarifying questions.
Ironically, AI is prompting us to revisit the mindset we should never have abandoned. It reveals who maintains good coding hygiene.
Lesson from Street‑Smart Coding
One of the lessons in Street‑Smart Coding (Chapter 3) is “Don’t rush to code.” This principle becomes even more valuable when AI serves as a coding assistant, emphasizing that context comes before code—with or without AI.