I am learning how to do vibe coding right. Vibe coding was a term coined by Andrej Karpathy for doing coding using LLM. And the world splitted with acknowledging it and flatly denying it.
There is a compelling reason why vibe coding is dangerous:
- LLM trained on a snapshot knowledge. Meaning, the updated knowledge might get passed out. This makes LLM produces old syntax that are no longer true. This is dangerous for building web where we are using the volatile Javascript libraries.
- LLM have the tendency to halucinate. This means it would generate codes that doesn't work.
They have fixed this with the new LLM specialized code. This was started by Antropic's Claude Code. And then, folks at Google releases the same with Gemini CLI. After that, Alibaba forks the Gemini CLI and built QWEN Code.
These are built either run in the project directory as a standalone, or run it as a companion with IDE such as VSCode. They have a distinctive feature as a specialized A.I. tools for coding:
- Their models are built towards populare programming language.
- They have a special markdown file (CLAUDE.md for Claude, GEMINI.md for Gemini, and any file for QWEN) that can be used to build the scope of the project.
- They can run the application they wrote to test the validity.
- They can call the web for affirmation and further knowledge.
GEMINI was released last month and it's improving well. For instance, in the earlier version, it only supports the authentication using GEMINI project. Now, they are using Google OAuth like other apps.
Expect them to fail. Expect them to burn your tokens. Because, they are sometimes do silly mistakes and learn from those mistakes. We sometimes need to step in and fix the problem ourselves or else the LLM will do the same mistake over and over.
Comments
Post a Comment