The article's claim that AI is responsible for a shrinking job market is tenuous at best. Economic factors are much more relevant. Rising interest rates make tech companies less attractive to investors, and this filters through companies as smaller hiring budgets.[0] And we're seeing relatively high interest rates compared to a few years ago. I am far from convinced that efficiency gains from AI correlate with—much less cause—software engineering job cuts.
This 1000%. As an investor, if interest rates aren't zero, why would you put your money in a risky startup when you could park it in a load of other places and still get a decent return? It's also not 2007 anymore, the low hanging fruit has been picked, most things that can be digitised and turned into services have been. You're going to have to come up with something extremely original and brilliant or utilising some kind of new tech in order to succeed on a big scale.
It's got very little to do with AI. Other industries are far more affected by AI than software engineering e.g in my opinion, the amount of graphic designers and copy writers are both going to shrink massively.
> It's also not 2007 anymore, the low hanging fruit has been picked, most things that can be digitised and turned into services have been
I think there's plenty of low hanging fruit left, especially when it comes to improving current systems that we assume inherently have to suck simply because they always have.
Plus there's new fruit growing every day. There's potentially huge opportunities in improving social media right now. Who would have guessed 5 years ago we could see legitimate vulnerability in the major platforms?
> It's also not 2007 anymore, the low hanging fruit has been picked, most things that can be digitised and turned into services have been
I would say this is somewhat true for "mostly software-contained" ideas, meaning things that don't take external domain knowledge besides software engineering
There is still plenty of things in multi-domain fields (software engineering + something else), like biotech. But being multi domain makes the barrier to entry MUCH higher for the main startup class of people (software engineers coming out of uni or big tech)
Threads, ActivityPub, Bluesky, nostr. There's hasn't been this much of a shake up in over a decade. Will it be enough? Who knows. But I'm hopeful we'll come out with something healthier than what we've been doing with algorithmic curation.
IMO the problem wasn't with how social media was implemented, it just coincided with a culture shift to being always online that the world wasn't really ready for.
These alternatives are fine (I'm a fan of mastodon myself) but I feel they don't really address the issues of the frayed social fabric we're currently experiencing in our societies.
In other words: these platforms aren't encouraging dialogue, they simply accepted that we can't get along so they started these small clubs so we can all have our carefully curated echo chambers.
So no, fixing social media isn't a low hanging fruit, not even close.
According to this Kurzgesagt video [1] the problem isn't echo chambers, the problem is the opposite. Encouraging online dialog isn't the solution if he's right. He suggests we go back to smaller online communities and more IRL interaction.
If anything I would expect a pop in demand for software engineers from ai. Those who know, know you can’t just ask chat gpt to integrate with the business by wiring a prompt.
Hi, can you please integrate your self into my business and increase my revenue by 20% and cut cost of workforce by 30%. Here are the master passwords to our aws account, jira, and payroll. Thanks! See you when in back from my 3 month vacation!
I doubt it'll affect graphic designers much. Artists, yeah. But i haven't seen any model or tool that is even remotely doing any meaningful graphic design solution or work. Majority of graphic designers have been using stock assets for graphics anyway, so ai will be another better asset generator.
AI may not be actually responsible for layoffs and hiring freezes, but there are CEOs using it as an excuse. It's getting irritating that journalists keep citing AI as the cause anyway.
They are the actual stochastic parrots. If anything, ai is replacing content writers. These people have not been providing value - the contrary, have been part of the problem - for years.
Another big factor (that doesn't get as much attention because it's wonkish) is the section of the TCJ Act of 2021 that forced software engineering to be amortized over 5 years instead of written off as an expense. Absolutely responsible for a huge pct of layoffs in 22 and 23.
Yep, we haven’t been allowed to hire anyone new on my team in a year. Definitely no headcount for junior engineers. With the cost of licenses for SaaS apps for a single developer my company would rather hire one senior engineer versus 2 juniors.
I never believed it. I remember telling people how this would turn out and getting kind of laughed at.
This industry is so fad driven. When something is the fad in tech it becomes heresy to question it.
I had a similar experience when everyone believed mobile would completely replace desktop/laptop systems. I explained why it wouldn’t, why it was for a different set of use cases, and was kind of scoffed at for that too.
The new generation has largely switched to exclusively using thier phones. I figure this will have some pretty negative consequences for computing, but was also inevitable given the state of things
People in the new generation who use computers to do real work don't use phones because you can't do very much in the way of real work on a phone. Developers also don't use phones because you can't develop (much) on a phone. They are pretty much pure consumption devices.
What's happened is that the phone has replaced the PC for casual consumer-level computer use, except maybe high-end gaming.
I’ve seen some crazy examples when things like Datadog become part of the stack or you have a really big “GitOps” operation with loads of SaaS things all chained together to deploy, audit, manage, etc. a build pipeline and a cloud deployment. Then there’s all the stuff like Zoom, Office, Google Workspace, and more that everyone needs.
It can get crazy fast, and once you let stuff into the stack it’s hard to get rid of it. This is particularly true when it becomes part of a process that is in turn tied up with compliance or is used by a lot of people.
Major source of “cost disease” creeping into our industry. I’ve taken to calling the overall phenomenon “low interest rate architecture.”
You may have a point there. I remember an exchange from a conversation I had with a recruiter a couple months ago where she talked about having to log in to literally 18 different systems to get her work done. Her company's solution was to implement SSO with Okta.
Okta is not a bad product by any means, and it certainly does work well. But, if SaaS costs are the problem, going from having 18 logins to remember to having a single login while paying for 19 SaaS tools seems like the opposite of a good idea to me.
Does anyone know to what degree the Section 174 changes are responsible for this market? It's been brought up separately in a few venues with dire warnings, but I've rarely seen it mentioned when people are talking about the current job market.
Before I switched to my current career, I was a PM at a pretty large software company that had a massive layoff in our US HQ and pivoted hiring almost entirely to Eastern Europe, Israel, and India in anticipation of this.
Nowadays we only hire a handful of strategic roles in the US - everything else R&D wise is abroad. Other peer companies in our sector did the same.
We had other considerations too (eg. Lack of RoI on US talent) but 174 was a driving factor
In my portion of the tech industry (Cybersecurity/Infra), there isn't enough talent in the US, so we'd end up paying a new grad roughly the same amount we could pay an 8200 veteran in Israel or a Flipkart security alum in India. The latter have more experience than the new grad, so we're getting RoI much quicker.
Not working as a PM anymore, but it is happening. Within 5-7 years most Cybersecurity PMs will end up being in Israel+India, along with Eng staff. All that will be left in the US is PMM and Sales Engineering.
This is the model the newer gen of Cybersecurity Startups are following (eg. Wiz, Cyera, etc).
I'm only answering in general, not personally, but this question is too vague. Would you take a job in your specialty in the US? If your answer is "yes", great! It's located in Nome, Alaska. Still interested?
There's a world of difference between living in Netherlands and Bulgaria, just like life in Honolulu is almost nothing like life in rural Alaska or NYC or rural Kansas.
I am US talent, and nope. It's tough for new grads, but not too bad for experienced US citizens.
Worst case I can work for the State or Federal government (especially now that all critical Tech roles have been classified as GS-14/15 now) and still earn more than as senior leadership in Europe.
Made a comment above. If SEs understood or could deign to get involved in politics at all, this would be reformed. It's 1) enormous 2) Poorly understood/ discussed
Yeah that's why all the programming jobs disappeared after 4GL languages made us more efficient, and then when intellisense happened the industry got even smaller. The real tragedy though was all these open source libraries.
Think about the (grossly simplified and somewhat questionable) historical availability of: electronic gates -> binary -> assembly -> C -> Python. Each of these steps as allowed one engineer to do the work of an entire team at the previous step.
And yet here we are: as technology has made engineers massively more productive, the need for engineers has gone up massively. Simply because the gains in productivity have made more projects doable.
So the question is not whether AI will make engineers more productive. The question is whether it will allow more projects to be undertaken. I don't have the answer to that question...
Are you using or have you used any of these AI tools for any length of time? Truthfully, no, I don't see it allowing 3 people to do 4 peoples' jobs, or for juniors to do the work of seniors. If I had to spitball a guess, I'd say there's potential for about a 10-15% productivity improvement on some tasks. Overall, as a senior/staff level software engineer, I can see these tools boosting my productivity maybe 5-10%. That's nowhere near "3 people doing the job of 4" levels.
Speaking of juniors doing the work of seniors, you've got it totally backwards. The people who benefit most from these tools are the more experienced, more capable people, not new grads fresh out of college. There are a ton of things a SWE does that no LLM will help you with -- at least not until virtual avatars get to the point where I can send fake me to a meeting and expect to still have my job at the end of said meeting. Being able to evaluate and understand the output of these tools is a key skill necessary for gaining a benefit from them, and that's precisely the skill that juniors, by definition, lack.
That said, this whole market dynamic of not hiring juniors at all right now is going to be a pain point for the industry at large in about 5 years if it keeps up. If you're a senior+ level SWE right now, that might end up being to your personal benefit, but it certainly won't benefit the tech industry when there aren't any mid-level engineers to hire, because we didn't cultivate, mentor, grow, and train them as juniors today.
Software isn't a finite problem that gets solved. We make what we can, given our resources.
Commercial software used to be painstakingly written in assembly, with little more than simple text editors. An average Python programmer today is 10x more productive than that, yet there are vastly more developers now.
People say that, but I have yet to work on a CRUD project. That's despite 15 years in industry and being involved with dozens of projects.
My guess is that simple CRUD use cases are automated via Google Spreadsheets or similar solutions. Same as there are off the shelf solutions for a simple online store, so no one is writing that from scratch any more. What's left is writing the complex systems with unique requirements.
For sure. One person I talked to said using a code assistant was just like having a smart and eager, but really green junior to help them, but without so much of the hand holding an actual junior needs, and none of the mentorship requirements.
I agree with you: it’s not that what you are saying is true, but that a lot of the people hoarding the money are salivating about this idea. For them, it’s easy to test the hypothesis: “let’s just fire people; with ai, at best we’ll become as productive as before or even better, at worst we’ll just have to hire the people again”. I hope that we won’t be riding the ai hype peak for too much longer.
I agree. I find it hard to believe that anyone who's actually studied and probed the limitations of LLMs and AI art generators actually believes that AI is going to take anyone's job anytime soon. That might have spurred some of the layoffs last spring, but, today, I absolutely do not believe it is.
That said, I have heard crazy stories from some people, like one CTO who was saying that "90% of code in [their] company would be written by AI" by the end of this year. And, there definitely has been a very palpable slowdown in hiring.
Let's put it this way: I applied for 50 jobs the first week of last month and got very little response. I've had three on-sites that didn't seem terribly difficult, yet no offer. I've had another 2 or 3 opportunities basically evaporate in front of me as the companies decided they were not going to hire at all, or changed their mind about what they were looking for. Two years ago, I might have applied to a total of maybe 30 over 2-4 months, and I'd have gotten probably at least 12-15 interviews, most of which would have probably turned into on-sites.
I'm starting to get more activity in my inbox now that the first of the year's rolled around. Let's hope it holds up long enough for me to land an offer :-)
But has anyone here seen any evidence of AI impacting software jobs? Maybe my company is just behind the times, but I'm seeing no indication that AI is going to displace anyone. Copilot and GPT-4 are convenient and speed up certain parts of the job, but the parts they impact aren't where the bulk of my time was spent anyway.
What exactly are you expecting? That a company will post a job opening and then tell the candidates "On second thought we've passed you all up for AI"?
That's never gonna happen.
As far as you seeing no indication, I saw a job posting today on Indeed for a Software QA and one of the bullet points included working with AI tools. So you may not see it anecdotally but it's out there.
I can't honestly say for sure, but I would say there's a possibility that AI was a factor in at least some of the layoffs last spring. Once June / July rolled around though, no way.
Generally speaking, people in charge shift a lot of blame for what may be pinned on them (interest rates in this instance) to other crises. Some of the manufacture of which can be traced back to said people. The case of Ai in this context is arguable. They did not create it, but are their hands clean of pushing Ai doomerism?
Agreed. And lacking apparent growth area is another reason. We had social media and rapid development of e-commerce in the 2000s, mobile revolution in the 2010s. What do we have now that sweeps the entire industry? Gen AI? But many companies are still trying to figure out what Gen AI can truly revolutionize.
Agreed, but Gen AI already has a lot of use cases, the challenges are with “taming” the technology - Rare hallucinations, RAG (or equivalent) figured out, more deterministic.
Assuming some of these challenges are solved, integrations will follow.
Imagine the web before CSS, JS, AJAX. That’s where we are with Gen AI.
No doubt Gen AI has lots of use cases. I'm just not sure if the use cases can lift the entire IT sector for the next N years like mobile or e-commerce used to do.
It could be that the expectation that AI will be able to replace tech jobs in the near to medium term (or at least lead to greater programmer productivity) is in fact influencing current C-level hiring decisions. Interest rates have plateaued and the consensus is that they could start dropping in the 2nd half of 2024.
Further, it's worth putting the components on npm individually and retaining 101, but making it depend on those modules (so it's really just a wrapper). This is what lodash should do imo. Lodash has the separate bits on npm, but lodash itself does not depend on them, which is a little weird.
I definitely did that for a while, but eventually found it annoying.
I know it's definitely better to package each of these as a
separate module. But uses of each of the tiny modules like these
can come and go very quickly and keeping the package.json up
to date becomes tedious..
I disagree greatly. I've used Angular in a team project that spanned several months and went into production at a startup. After reading this article, I realize how much of my experience is reflected in it. I personally experienced each of the problems on his "bad parts" list, except for the name clashes issue (#3). For the last few months I've been telling people that Angular is powerful, but it has an extremely steep learning curve. I've been reluctant to tell people to try it because, why you can do a lot with it, you must baptize yourself in the framework before you can write anything half-decent. It's a little depressing to see someone with deeper experience come out and say it does not get better. But I know that as much as I want to like it, Angular imposes too much for me to actually recommend it.
So, this article is definitely not generic. Maybe, if you're reading at the bottom of the article, you can call it generic, because the advice for future projects is good. But otherwise, it's spot on. I can only hope that the Angular dev team sees this blog post and incorporates the recommendations into version 2.
I'm really excited about this kind of service. People have been moving increasingly online in everything they do, and music is no exception. I believe that in the future the vast majority of media consumption will use streaming. (As if it isn't, already!) My main side project involves this kind of service, so I know at the very least it can be complicated to write a server/website which can chomp diverse APIs. [1] What puzzles me is how this could be monetized--I'm sure YouTube and SoundCloud aren't in a hurry to stream their media for free without seeing license fees.
[0] https://www.nytimes.com/2023/01/23/technology/tech-interest-...