For the first time ever I did some vibe coding this week. I needed what amounts to putting a file in a folder and watch until a file in another folder appears, with a monitoring utility.
I was very impressed with the definition phase. It was a back and forth about the spec until it was what I wanted. I then let it built.
This was quite interesting. It set up a scaffold, wrote tests and ran the tests, fixing errors as it went, then went on to finalize the app. Magic.
Finally it was done, so I go in to play with it. Riddled with bugs. Try to get it to fix them. More bugs. End to end spend a morning doing all this until I gave up and manually wrote what I need in about an hour.
Conclusion: Vibe coding is a complete waste of time. Worse: in the wrong hands it is dangerous. You need to be a pretty damn good programmer to assess the output, and if you are, why the fuck would you use it in this manner?
It reminds me a bit about the no-code/low-code evangelism of some years ago. To the novice it looks magical and a world of possibilities, and the road to riches. But in reality you can only use those things if you have decent programming skills, otherwise it’s a path nowhere.
Need a quick demo? Sure, this might help the novice. Want something for production? Get a professional.
The thing about low code is the successful products in that field have their blocks built by experienced teams. I’ve heard of setting up low code apps via LLMs and that almost makes sense. They can only do as much damage as a bad project manager cosplaying a solution engineer, scrapping the whole exercise isn’t too bad, and they can be a nice demo for the client.
For the first time ever I did some vibe coding this week. I needed what amounts to putting a file in a folder and watch until a file in another folder appears, with a monitoring utility.
I was very impressed with the definition phase. It was a back and forth about the spec until it was what I wanted. I then let it built.
This was quite interesting. It set up a scaffold, wrote tests and ran the tests, fixing errors as it went, then went on to finalize the app. Magic.
Finally it was done, so I go in to play with it. Riddled with bugs. Try to get it to fix them. More bugs. End to end spend a morning doing all this until I gave up and manually wrote what I need in about an hour.
Conclusion: Vibe coding is a complete waste of time. Worse: in the wrong hands it is dangerous. You need to be a pretty damn good programmer to assess the output, and if you are, why the fuck would you use it in this manner?
It reminds me a bit about the no-code/low-code evangelism of some years ago. To the novice it looks magical and a world of possibilities, and the road to riches. But in reality you can only use those things if you have decent programming skills, otherwise it’s a path nowhere.
Need a quick demo? Sure, this might help the novice. Want something for production? Get a professional.
The thing about low code is the successful products in that field have their blocks built by experienced teams. I’ve heard of setting up low code apps via LLMs and that almost makes sense. They can only do as much damage as a bad project manager cosplaying a solution engineer, scrapping the whole exercise isn’t too bad, and they can be a nice demo for the client.