But if you're aiming to provide the best education to future professionals so that they are able to excel at their field, you'd serve them far better if you taught them how to effectively use Wolfram Alpha.
The truth of the matter it that the bulk of the work in software development, and what makes more economic sense, lies in growing a system through accretion and adopting higher level components developed and maintained by third parties.
Nonsense. MIT engineers don't waste their time thinking about assembly instructions when they are tasked to design a distributed system. Competent engineers know how to switch to high-level concepts where they make sense.
The difference between trade schools and universities is in the depth of fundamentals. Always has been, still very much is.
It's definitely nonsense. You're failing to understand that today's "fundamentals" are not the parlour tricks you have in mind. You're in a better position to get ahead in providing software development services if you're able to put together a customized script that launches traefik, nginx, rabbitmq and a few nodejs services fully dockerized than if you're able to tell what opcodes your compiler generated out of C code.
I agree. The problem is that you have a gross misconceptions about what "strong foundation" represents in this day and age.
I knew guys who fail to stop and think about "foundations" and instead use that as a scapegoat to shoehorn their gatekeeping logic and as a form of ladder-pulling. It's the kind of character who think that playing trivia games is a good way of asserting whether a candidate is a competent engineer.
And to internalize the principles you have to at least use them in a didactic setting.
I asked him how it was going, and he said it was going well. But that he was confused about some things.
I asked them what they were.
He said, paraphrasing, "What's the difference between disk and RAM?".
On the one hand, it's great that someone without such fundamental understanding was able to accomplish something. On the other hand, it was a bit disturbing.
I've run in to several folks who do not understand first principles, and some times are combative when I explain how I "know" something "can't work" or "didn't happen". I'm sure we've all heard those things. "What about <absurd thing>?" "No. Just...sigh...no."
So maybe their confusion is less bad than it seems at first? Maybe?
I recently was with a group of dev ops folks that had written a 100k of terraform code. It worked great, nice improvements for consistency of cloud deploys, and decreased time to implement and all that.
Problem was when terraform would implement breaking changes in new versions. We had so much code to fix, the group just wanted to stay back at version N - 1. But I knew where that would end, in five years they would have buggy unsupported old terraform needed some big heroic project to upgrade, with risk and so on.
So I wrote a simple Python script that was a recursive descent parser/transformer patterned on expat with callbacks. It was able to transform all the code and bump the terraform version at the latest version.
It took a while to write, but was tremendously fun, gave us a powerful tool for transforming and therefore owning the 100k+ LOC of terraform, which other wise threatens to become a congealed unmaintable blob.
Even writing custom off the shelf services can have big payoffs, in efficiency or in operational ease. There are two ways in which the body of existing software is full of good solutions; there are a ton of turn key solutions, from Hadoop to NoSQL to Airflow to the other fifty nine Apache projects, Kafka, etc. etc. etc.
Also, modern languages have so many excellent facilities for highly scalable, cheap and reliable software that is shall we say designed precisely o meet a given customer need. The custom solution can be faster better and cheaper than the turnkey open source system, which usually has a high cost of complexity and a high cognitive burden for ops to understand the configuration space and so on.
Maybe we all do not have to know first principles. But somebody does.
And if MIT won't teach first principles, who will?
The idea of genius men making stuff from first principles is just some kind of Hollywood schlock version of programming anyway. It's the stuff you see on NCIS when they make an episode about hackers. All the good stuff has been a team effort, even if it starts out as a solo project. Maybe with a few exceptions like cURL.