This will create what I am calling a “seniority crunch”. Entry level jobs are hard to get, humans aren’t learning the deeper aspects and reasoning parts of knowledge work. Gap created for people with this knowledge. Senior IC jobs demand goes up. Fewer with xp to fill it. LLMs don’t reason well. They can’t do that and likely never will be coaxed into it. Just one scenario I think is plausible.
No, this is the likely path. LLMs are hitting a wall. They’ll get better, much better, but high context “brownfield” tasks are going to remain difficult and require human handholding.
It's not like juniors roles will go away. More and more - it's looking like there's a higher bar for high compensation that was the norm. It seems like juniors SWEs will have to toil more to reach higher level of skills, become valuable to their teams, before they "make" it. I'm seeing this trend across the engg leaders I speak with.
I think this is the most likely outcome. It'll drive salaries up in 5-10 years when there's a serious talent pipeline shortage and the demand for senior talent isn't satisfied by a dry pipeline.
Or it will lead to demand for massive government job programs, huge tax increases to pay for, free enterprise being smothered, everything being run by government bureaucrats.
All thanks to AI entrepreneurs, such a great achievement. Just like AirBnB destroying the historical centers nowadays, disruption at its best - just a different disruption from what they claim.
Or it will require college students to complete one or more internships which use AI to show they have already learned to be productive using AI before they graduate.
It's not an AI issue. It's that in 4 years of a CS degree, CS students never touch a single kubectl command or barely build one functional web application in one software engineering related course. It's the failure of CS programs that is causing job market issues for college graduates.
It's just that when money was easy, companies could pay for training interns and recent grads things they should have learned in school. Now money is tight, so we see these job market problems. I am hopeful that colleges will adapt their curriculums based on the changing job market.
> It's the failure of CS programs that is causing job market issues for college graduates.
Definitely not. These failures have existed for decades. But it ultimately doesn't matter. You have the capacity to learn or you don't. CS programs are a waste of time, sure, but they have nothing to do with the job market.
I'd go so far as to say your comment is a perfect example of the problem. "kubectl" doesn't mean jack shit to being a good engineer. It's a technology we will use for a few years and then move onto something else.
Absolutely nobody is cutting junior positions because they don't know "kubectl"
The problem is developers are now DevOps, QA, PMs, Customer Support, and everything else.
CS programs teach fundamentals, though not necessarily the most popular tools in the industry. Those skills help regardless of whatever specific tools you end up using. I don’t think it’s a waste of time.
> The problem is developers are now DevOps, QA, PMs, Customer Support, and everything else.
As long as the hour expectations are reasonable I actually like this. This teaches you so many more skills than being pigeon-holed into software engineering. Granted, I like the idea of one day founding a business so I may need to be able to do all of these rolls some day.
I think you may be taking the parent comment too literally. Kubectl itself doesn't matter, replace it with anything you want - it represents the lack of the graduates ability to do anything useful in the real world.
Is doesn't make much sense to spend four years preparing for a job (market) and be completely unprepared at the end of it.
Well being a good engineer doesn’t mean “Jack shit” if you’re homeless and hungry because no one will hire you to exchange money for labor to support your addictions to food and shelter.
If I have to explain everything in detail and check behind your work as a junior dev snd wait for you to get up to speed, I might as well just use ChatGPT.
If I do need a perdón, why would I hire a junior engineer when I can just recruit an underpaid mid level developer more whose already proven himself and gotten up to speed especially if I can hire someone cheaply living in MiddleOfNowhere Nebraska who is willing to work for less than I made over a decade ago?
Implicit in this is an assumption that the purpose of a higher education is to give graduates the skills demanded by the businesses. What if it (the purpose of a higher education) was something else? What if it had always been something else, and only in the last few decades did a certain segment of the population try to convince us that the only reason higher education existed was to churn out workers that could slot straight into entry-level jobs without businesses having to invest in any training first?
College is a massive time and monetary commitment. The reason the vast majority of people go through it is because employment opportunities are gated away behind a degree. Without one, finding the job you want can be extremely difficult, and at times even impossible.
It always bothered me how colleges hold someone's future career hostage like this, forcing them to go through them, and then they pretend it's not the case and that college is simply intellectual enrichment.
The gating that colleges do also makes it much more difficult to change careers than it should be. The whole system does the task it's given extremely poorly, and pretends that it's not even responsible for the task. Great, fine. Let's work on alternative ways to handle job credentialing then.
It's not a "certain segment" that changed their mind. It's the reality of what's happening to our country (countries). We can no longer, at large, afford the luxury of spending so much money ONLY to "learn to think". We also need practical education more than before. That's the undeniable nature of what has happened to the economy, especially in software.
I can guarantee you that the only people who went to college to be “better citizens of the world” were those who were privileged enough to have parents who could afford to subsidize their living.
And thank goodness we don't teach kubectl and never will.
Knowledge at the level of kubectl is worthless to the vast majority of grads. It's also extremely time-limited. In 5 years everything will have moved on.
Web apps? What fraction of our grads build web apps? How much will the technology for doing so change in 5 years? 10 years?
Universities are not vocational schools. We teach people to think. Then they learn what the details are on the job. This has always been the role of universities.
If you want something else, go to a coding bootcamp.
This isn't viable. This isn't a normal vocational job - it's too abstract/overwhelming for everybody to learn all of it on the job to the needed degree. We need to stop tricking kids into this path. Either CS needs to teach practical skills (I understand things evolve fast, but you're exaggerating, and regardless, an education in the current thing is absolutely invaluable, and also prepares you for learning the next thing on your own), or we need to make it much more clear that another education path is better. Also, boot camps are worthless for this.
Universities are research oriented, kubernetes is a practical skills which you'd expect to learn at a trade school no different to learning fluid dynamics vs plumbing
I have never met a single human being in person who did not believe that a crucial part of acquiring practical skills in software (not "trades") for young people is to go to university. Maybe people shouldn't theoretically expect this, but there is theory, and then there is reality. We need to make them match, in whichever direction.
If teaching fundamentals, then force students to utilize these fundamentals to build into applied or industry usecases.
For example - leveraging the K8s thread - force students in their OS class to understand HOW to apply data structures to manage scalability problems AND forcing students to understand Linux kernel internals like cgroups.
Most universities don't offer course to connect these fundamentals together - not even my Ivy League alma mater based on my own survey of curricula.
Everyone is incentivized to understand the bare minimum, overindex on Leetcode, and skim over harder fundamental courses. On top of that, ime at my alma mater, most CS faculty was essentially applied math nerds who could see beauty in a well formatted inductive proof but would glaze out when pushed on implementing Paxos using best practices around kernel or system performance. And vice versa for the system nerds.
In a world where AI/ML can increasingly automate away boiler plate work and even conduct limited reasoning, understanding how fundamentals and (shudder) first principles are connected into become a product or solution is what matters.
And thus, this becomes a critical thinking problem, which just cannot be learnt without experimentation and getting into the weeds.
Whey he’s advocating for is that people don’t want to get in tens of thousands of debt without expecting to get a job because they need money to survive and pay off said debt
This has not been the case the last 10 years in most of the top 50 programs. Capstone projects and senior seminar classes often touched upon more "serious" topics. I've seen classes focusing on projects related to distributed/cloud computing, complex web apps, or then often hot-topic of the year technologies (game AI, computer vision, machine learning)
Aside from a few exceptionally awful periods like 2009, when did fresh CS graduates struggle to find tech jobs? We all went through similar curriculums and picked up industry knowledge perfectly fine.
Careful about your labels! I went to brown in the late 90s and considered a cs degree before moving to math.
Even then among academic programs brown was considered a bit low brow in that it spent a fair amount of time on software and systems engineering: this was all part of the cs department, but it wasn’t science per se.
Today most cs programs are in fact software engineering programs; well and good, but they give themselves the same name as the original meaning which was, while engaged in engineering, more theoretical.
I can think of so many counter examples to what I just wrote I’m sort of excited for the comments, but tldr: in the late 90s, if you wanted engineering skills you hired MIT, Caltech, not Harvard. (Stanford alums didn’t move east and even then were not generally hirable for cash)
I’ll note my current perspective is probably aligned with your complaint - we need more engineers and quality engineers than we do theoretical cs undergrads by at least two orders of magnitude. The best schools taught their engineers theory, and I think that goes with the history of the discipline - most of the greats were tinkerers at least, more usually engaged in real engineering work while working on theory. The other way: getting theorists to become great coders - seems less common.
And in fact my era at brown produced a number of influential engineering folks like brian cantrill, so despite its non exalted status, it did the world some good!
And yet if you flip the paradigm entirely, you pretty much get coding bootcamps, which certainly don't have a great track record either. The answer is probably some more ideal balance between theory and practice, like Waterloo's CS program.
If AI is really wrecking the job market structurally due to new efficiencies, one would hope those effects will be permanent (i.e we won't need juniors anymore because AI will fill those spaces permanently, with or without human intervention; and thats not going to change because AI is only going to get better, blabla, yada, yada).
If it is not, it would be a matter of eventually reaching AI winter status and/or just finding a new thing to hype and the tech job market will open its gates again.
Of course there is some nuance to it, both things can happen at the same time with different intensities.
I just haven't seen this. Not at all. AI is a useful tool, but it is nowhere close to replacing humans, except where companies want to be bastards by forcing customers to use an AI instead of a person. It's going to be quite a while until an AI actually takes a full job.
Money is tight for humans. While the sea level has gone down for everybody over the decades, businesses have not shared a proportional level of financial suffering compared to humans, by a long shot.
...say the businesses pulling in record revenue and profit numbers.
Do the people who say things like "money is tight" think most of the rest of us are stupid? Based on how many people parrot those lines despite all evidence, maybe we are communally that stupid.
Readit News
The industry is cutting itself off at the knees.
It's just that when money was easy, companies could pay for training interns and recent grads things they should have learned in school. Now money is tight, so we see these job market problems. I am hopeful that colleges will adapt their curriculums based on the changing job market.
Definitely not. These failures have existed for decades. But it ultimately doesn't matter. You have the capacity to learn or you don't. CS programs are a waste of time, sure, but they have nothing to do with the job market.
I'd go so far as to say your comment is a perfect example of the problem. "kubectl" doesn't mean jack shit to being a good engineer. It's a technology we will use for a few years and then move onto something else.
Absolutely nobody is cutting junior positions because they don't know "kubectl"
The problem is developers are now DevOps, QA, PMs, Customer Support, and everything else.
As long as the hour expectations are reasonable I actually like this. This teaches you so many more skills than being pigeon-holed into software engineering. Granted, I like the idea of one day founding a business so I may need to be able to do all of these rolls some day.
I think you may be taking the parent comment too literally. Kubectl itself doesn't matter, replace it with anything you want - it represents the lack of the graduates ability to do anything useful in the real world.
Is doesn't make much sense to spend four years preparing for a job (market) and be completely unprepared at the end of it.
If I have to explain everything in detail and check behind your work as a junior dev snd wait for you to get up to speed, I might as well just use ChatGPT.
If I do need a perdón, why would I hire a junior engineer when I can just recruit an underpaid mid level developer more whose already proven himself and gotten up to speed especially if I can hire someone cheaply living in MiddleOfNowhere Nebraska who is willing to work for less than I made over a decade ago?
It always bothered me how colleges hold someone's future career hostage like this, forcing them to go through them, and then they pretend it's not the case and that college is simply intellectual enrichment.
The gating that colleges do also makes it much more difficult to change careers than it should be. The whole system does the task it's given extremely poorly, and pretends that it's not even responsible for the task. Great, fine. Let's work on alternative ways to handle job credentialing then.
Knowledge at the level of kubectl is worthless to the vast majority of grads. It's also extremely time-limited. In 5 years everything will have moved on.
Web apps? What fraction of our grads build web apps? How much will the technology for doing so change in 5 years? 10 years?
Universities are not vocational schools. We teach people to think. Then they learn what the details are on the job. This has always been the role of universities.
If you want something else, go to a coding bootcamp.
If teaching fundamentals, then force students to utilize these fundamentals to build into applied or industry usecases.
For example - leveraging the K8s thread - force students in their OS class to understand HOW to apply data structures to manage scalability problems AND forcing students to understand Linux kernel internals like cgroups.
Most universities don't offer course to connect these fundamentals together - not even my Ivy League alma mater based on my own survey of curricula.
Everyone is incentivized to understand the bare minimum, overindex on Leetcode, and skim over harder fundamental courses. On top of that, ime at my alma mater, most CS faculty was essentially applied math nerds who could see beauty in a well formatted inductive proof but would glaze out when pushed on implementing Paxos using best practices around kernel or system performance. And vice versa for the system nerds.
In a world where AI/ML can increasingly automate away boiler plate work and even conduct limited reasoning, understanding how fundamentals and (shudder) first principles are connected into become a product or solution is what matters.
And thus, this becomes a critical thinking problem, which just cannot be learnt without experimentation and getting into the weeds.
The pandemic likely also caused further changes in industry and their requirements.
Even then among academic programs brown was considered a bit low brow in that it spent a fair amount of time on software and systems engineering: this was all part of the cs department, but it wasn’t science per se.
Today most cs programs are in fact software engineering programs; well and good, but they give themselves the same name as the original meaning which was, while engaged in engineering, more theoretical.
I can think of so many counter examples to what I just wrote I’m sort of excited for the comments, but tldr: in the late 90s, if you wanted engineering skills you hired MIT, Caltech, not Harvard. (Stanford alums didn’t move east and even then were not generally hirable for cash)
I’ll note my current perspective is probably aligned with your complaint - we need more engineers and quality engineers than we do theoretical cs undergrads by at least two orders of magnitude. The best schools taught their engineers theory, and I think that goes with the history of the discipline - most of the greats were tinkerers at least, more usually engaged in real engineering work while working on theory. The other way: getting theorists to become great coders - seems less common.
And in fact my era at brown produced a number of influential engineering folks like brian cantrill, so despite its non exalted status, it did the world some good!
Deleted Comment
But seriously, it's both.
Also, CS? That's not a degree for building apps, right? If it is, there's definitely a failure to make that clear.
If AI is really wrecking the job market structurally due to new efficiencies, one would hope those effects will be permanent (i.e we won't need juniors anymore because AI will fill those spaces permanently, with or without human intervention; and thats not going to change because AI is only going to get better, blabla, yada, yada).
If it is not, it would be a matter of eventually reaching AI winter status and/or just finding a new thing to hype and the tech job market will open its gates again.
Of course there is some nuance to it, both things can happen at the same time with different intensities.
...say the businesses pulling in record revenue and profit numbers.
Do the people who say things like "money is tight" think most of the rest of us are stupid? Based on how many people parrot those lines despite all evidence, maybe we are communally that stupid.
For Some Recent Graduates, the A.I. Job Apocalypse May Already Be Here
https://news.ycombinator.com/item?id=44146650
And conversely:
The ‘white-collar bloodbath’ is all part of the AI hype machine
https://news.ycombinator.com/item?id=44136117