Salvatore Sanfilippo's (somewhat dry, in parts) piece on using LLMs as a coding helper matches my experience. I too prefer ChatGPT collaborating from its sandbox vs a helper sliding less predictably into my editor's autocompletes; and it is so very useful as a kind of "self-customizing reference", able to take on some medium complexity challenges but failing on more intricate system (especially when iterating on even its own code, I've found; it can lose track of the original specifications all too easily)
One interesting paragraph was as much about the history of the programming industry as much as it was about LLMs:
In the field of programming, perhaps their ability would have been of very little interest up to twenty or thirty years ago. Back then you had to know a couple of programming languages, the classic algorithms, and those ten fundamental libraries. The rest you had to add yourself, your own intelligence, expertise, design skills. If you had these ingredients you were an expert programmer, able to do more or less everything. Over time, we have witnessed an explosion of frameworks, programming languages, libraries of all kinds. An explosion of complexity often completely unnecessary and unjustified, but the truth is that things are what they are. And in such a context, an idiot who knows everything is a precious ally.
I think that's one of the great dualities in the UI coder industry, right up there with holism vs reductionism: would you as a coder prefer to make something novel in an old-hat framework, or recreate some common interaction using the new hotness? The latter view seems to have won out - there's a lot of neophilia, with the assumption that the leading edge is doing important work that outweighs the benefits of familiarity.
And of course that vast footprint of possible systems to have POSSIBLY worked in seems to make current jobhunts that much tougher (anyone need a strong, veteran UI Engineer with deep React and TypeScript? ;-) ) - those few companies who aren't in hiring freeze (or at least slush) can take their time and look for their unicorns that match.
No comments:
Post a Comment