As developers increasingly lean on AI-generated code to build out their software—as they have with open source in the past—they risk introducing critical security failures along the way.
Vibe coding works when you need to say connect to some API and can feed the model a bunch of docs.
It’s great for very low skill, low maintenance, low risk code that I can easily and reliably regenerate.
Increasingly coding models are improving at architecture choices, Claude 4.5 vs 4 is way better here. But ultimately it’s inferior to a ginger making those choices.
It’s also a great debugger and reviewer.
I used it this weekend to connect to an API and to build a table of constants by just feeding it docs. That was a huge time saver.
I also used it to try and implement stuff and I gotta say once it hit tricky things it started trying to game it and just say it works.
Vibe coding works when you need to say connect to some API and can feed the model a bunch of docs.
It’s great for very low skill, low maintenance, low risk code that I can easily and reliably regenerate.
Increasingly coding models are improving at architecture choices, Claude 4.5 vs 4 is way better here. But ultimately it’s inferior to a ginger making those choices.
It’s also a great debugger and reviewer.
I used it this weekend to connect to an API and to build a table of constants by just feeding it docs. That was a huge time saver.
I also used it to try and implement stuff and I gotta say once it hit tricky things it started trying to game it and just say it works.