In the very same way as you walk or swim. We don’t think “well, ok, left foot forward, now carrying the weight of body weight, now right foot forward,” but we also don’t breaststroke while on the walk.
I could not agree more. Terminology does not matter at all. I was writing about “being familiar with approaches,” not about “knowing all of them by name.”
Fair enough. honestly I am probably just permanently scarred by a colleague responding to my challenge about introducing a global state via redux for a hyper-local concern by saying simply “it is more declarative”…
I think the post is written in a toungue-in-cheek kind of way, and people here take that too seriously.
I enjoyed it. And we more or less did the X langs in X weeks in a uni course and it blew my mind. I agree everyone would benefit from knowing more than 3 as an intellectual exercise at least. And they should be FP, OOP, and Logic. The OOP, Reactive, and Aspect combo doesn’t stretch your mind in the same way.
With LLMs I feel the barrier to introduce another language into an org is now lower. Whereas before I would be against adding more languages into an org because of future tech debt/hiring.
Yeah, I avoid talking seriously even to my lawyer, and I’ll never stop wondering how much people are willing to be offended instead of smiling.
This I could not disagree more. The main danger posed by AI is us less understanding our own code. LLM is great to help us to write some routine code in the language / domain we know, because we can easily validate it. Allowing LLM to write the code in the language / domain we cannot validate is extremely hazardous.
Consider you want to please me writing smth in Catalan (or even Russian on that matter,) and you simply ask your favorite translation service, “How do I say ‘Great job!’,” and it comes out with “Gran feid!” instead of “Gran feina!”—and you send it to me because there is no chance you can tell “feina” (“job”) from “feid” (“sh*t”.)
I think onboarding speed onto new languages is 100 times faster now. Sure there is a danger of not reading or comprehending the solution, but that was already a danger with Stackoverflow answers. Syntax is not an obstacle anymore. Asking the LLM to write an idiomatic solution, then question each line until you understand it. If it smells wrong ditch it or tell the LLM to rewrite it with the new angle.
And if someone just needs to add a new clause to something existing I think it is much easier to do “relatively” safely now.
If your code is responsible for lives or billions of dollars.. sure. Don’t do that. Bust most SaaS is not like that. I bet management would rather have speed over a minor error that can be fixed if when encountered.
C’mon. We’re breaking spears in discussions about how we can get developers to pay any attention to pull requests in the modern era, and you’re honestly saying that people will figure out line by line what LLM has generated for them? Until they understand it? I thought I am a naïve dreamer, but you’ve completely surpassed me on that one!
Management always prefers that but they never see that gaining 10% velocity this week loses 30% velocity the next month. I’ll side with @mudasobwa here: we the people still genuinely bring value by having full mental context of the code (as much as our brains allow it). If we give up that as well then we become fully redundant.
I’d say how many paradigms a good developer should know is a number that will defy empirical quantification. It’s the same sort of issue as how one can tell if one is dealing with a “good” developer or not.