This is why nearly everything looks like a one weekend pet project by the standards of software engineering.
My claim is supported by the post article and many points there, for example. Another example is my own experience working with python ecosystem and ai/ml libraries in particular. With rare exceptions (like pandas) it is mostly garbage from DevX perspective (in comparison of course).
But I admit my exposure is very limited. I don’t work in ai area professionally (which is another example of my point btw, lol))
Do you perhaps mean small language models?
I doubt Llama or Deepseek were vibe coded..
I would use those as examples of an exception from my generalized point.
Anything else? Just a handful of tools you can call professional of thousands and thousands used everyday?
A lot of AI work is done by people that "dash-shaped" -- broad, but with no depth anywhere.
Then there's a few I-shaped people that drive research progress, and a few T-shaped people that work on the infrastructure that allows the training runs to go through.
But something like a protocol will certainly be designed by a dash, not an I or a T, because those are needed to keep the matrices multiplying.