Vibe Composing (Pt. 2) - The "boring" but exciting parts
After some back and forth with learning and reading all I needed for the project I finally dipped my hands into installing things and creating my project. This was a wild ride because I asked A.I to create a python project for me. The wild ride begins!
![]() |
Photo created with Google Imagen |
The amazing triviality
So I asked the LLM to create a Python project and use the UV package manager. UV was not installed and I had to install it. Unlike Cursor which would have installed it for me CoPilot did not, just gave me the command and I applied it myself. And then the picture below happened.
![]() |
Screenshot from my own machine |
The project was created and it even "read my thoughts" and what I meant with RAG and added the relevant description. At this point my mind is blown. Obviously previous context was responsible for the description and yet it was pretty amazing.
The rest of the project was completed mostly with completions because I had to practice my Python skills. However even for that the AI in the IDE was very helpful.
The final act was to update the project.toml file
![]() |
Screenshot from my own machine |
Final Thoughts
Vibe coding and in fact composing is all the buzz, or is it? As someone pointed, we who try these new tools may be living in an echo chamber and see only vibe coding around us. Be that as it may, more and more tools leveraging A.I agents come out every day (cline, roo code and taskmaster to name a few) and the future of coding will inevitably be governed by coding agents supervised (at least at first) by human developers.