Not to fully argue against your point, but I do want to push back on the citations bit. Given the way an LLM is trained, it’s not really close to equivalent to me citing papers researched for a paper. That would be more akin to asking me to cite every piece of written or verbal media I’ve ever encountered as they all contributed in some small way to way that the words were formulated here.
Now, if specific data were injected into the prompt, or maybe if it was fine-tuned on a small subset of highly specific data, I would agree those should be cited as they are being accessed more verbatim. The whole “magic” of LLMs was that it needed to cross a threshold of data, combined with the attentional mechanism, and then the network was pretty suddenly able to maintain coherent sentences structure. It was only with loads of varied data from many different sources that this really emerged.
Maybe I’m an old fogey, but I usually hear more pushback against visual languages as being too finicky to actually create anything with and I usually advocate for a blending of them, like working in Godot and having nodes to organize behaviour but written scripts to implement it.
I really appreciate the talks from Bret Victor, like Inventing on Principle (https://youtu.be/PUv66718DII), where he makes some great points about what sorts of things our tooling, in addition to the language, could do to offload some of the cognitive load while coding. I think it’s a great direction to be thinking, where it’s feasible anyways.
Also, one reason folks new to programming at least struggle with text code is that they don’t have the patterns built up. When you’re experienced and look at a block of code, you usually don’t see each keyword, you see the concept. You see a list comprehension in Python and instantly go “Oh it’s a filter”, or you see a nested loop and go “Oh it’s doing a row/column traversal of a 2d matrix”. A newbie just sees symbols and keywords and pieces each one together individually.