Alignment: Keeping the AI on Your Team
Definition: The extent to which an AI’s goals are in line with its creators’ goals, ensuring safe and beneficial AI behaviour.
What is Alignment in Coding?
In broad AI safety, Alignment means “don’t destroy humanity.” In Vibe Coding, Alignment means “don’t destroy my repo.” It’s about ensuring the AI adopts Your Style, Your Conventions, and Your Goals.
- Misalignment: You use React Hooks. The AI generates Class Components.
- Alignment: The AI reads your
.eslintrcand generates code that passes the linter on the first try.
Techniques for Alignment
- The
.cursorrulesFile: This is the most powerful alignment tool in vibe coding. It’s a system prompt that lives in your repo.- Example: “Always use TypeScript. Prefer arrow functions. Use Tailwind for styling.”
- Few-Shot Prompting: Show, don’t just tell. Paste an example of “Good Code” from your project before asking for new code.
- Feedback Loops: When the AI messes up, tell it “Don’t do X, do Y.” Modern tools remember this preference for the session.
The “Vibe” IS Alignment
“Vibe Coding” is essentially achieving high-bandwidth alignment. When the AI “gets the vibe,” it means it is highly aligned with your mental model. You only need to say “Fix it,” and it knows how you want it fixed.
Future of Alignment
We are moving toward “Personalized Foundation Models”—small fine-tunes of models like Llama 3 that are specifically aligned to you and your company’s proprietary codebase.