Coding is cool and I always had fun with it.
Becoming a great developer is hard. I remember to many waves learning technologies, going deep on internals and try to understand the limits of what I can optimize.
Today, vibe coding is really challenging us to keep code sane and sanitized.
Too long rows, too many things, too dumb things. Sometimes I feel it’s really spoiling my work and I need to come back on it.
I like the experience I can have and speed up the way I learn new things but vibe coding is really not a thing for the day as a developer.
I’m using the copilot to:
- Resume and walk me through key code to understand some patterns and communication between objects.
- Help me to write and sketch new tests for my implemented classes.
- Discover and generate data for my database.
- Help me learn c# in the speed of light.
- Help me to setup benchmarks isolating components to check performance.
What I gave up on using LLM:
- Drive the logic until “it works”
- Set the basic components
And, I’m back on the terminal. Using iTerm with neovim. Freely writing without over promotional explosion of words everywhere.
I love using AI, I continue vibe coding some projects, but I decided that I need to find a more sane way that makes me think and act, without proxying defaults to LLM just for the sake of saving time.
I see more and more this is just a cognitive load transfer that we’re being abusively consuming for humans, making the advancement fast for AI and the learnings heavily productized by them. In exchange we see some new scaffolds of our ideas popping up in the speed of the light. Where you’d never touch, now you just vibecode it.
Things I loved I explored with AI last few weeks:
- I had a great experience getting to touch some Rust code and work on a pluging for PGDog which is a project I admire a lot.
- I recently created the dome builder to learn more math around geodesic domes and perfect cuts
- I revived the mandala playground from my Art Therapy degree. I loved working on this projects.
- I’m working on the dj-data as the first UI simulator/core audio engine for my sonification device.
- I’m building a Raspberry PI project to stream sound from my home to listen and birds from their sounds.
- I rebuild the Postgresql Internals Workshop as a web material here.
It’s being fun and interesting. I like doing these things.
What I’m getting more and more concerned is that I could just go and paste this entire text I fully wrote by hand now on LLM to generate more noise for you. To make it brilliant oustandingly noised from the average of any topic. Just feel the trap.
It’s cool to use for fun, it’s cool to learn and get more words and a better domain about any topic, but be careful to not carry this noise on.
Communication is a priceless asset that now is really turning into garbage. I decided to get back to my old-school mode on my terminal. Write every word and think about every sentence. For now I decided to respect the reader and keep my personal blog pieces content original.
