Why do some people have a shit time with LLMs for programming while others love it?
What the latter group does is create tons of scaffolding to properly restrict LLMs, and acquire new skills to cope with all the unpredictable stuff they might do.
And because people are amazing at adapting the tools theyβre given and totally underestimate the extent to which they do it, this is mostly invisible.
In this post, I argue that the amount of skill and scaffolding we grow is an incidental consequence of how badly AI coding tools are designed.
If we spent less time congratulating machines for the work people do and we instead supported people more directly, we could get much better outcomes.
https://ferd.ca/the-gap-through-which-we-praise-the-machine.html