The easiest explanation for the whole chart is "Software dev is more reactive to the monetary policy environment than most other industries, because more of it is funded by new capital investment -- whether VC money or otherwise -- instead of ongoing operations of stable firms compared to most other industries."
Trying to add other explanations is fun, but there's not a lot of evidence for any of them, or even that more explanations are needed.
You mean the change from deductible as a current expense to five-year amortization under changes to Sec 174? In an easy money situation, the way it increases tax liability in year one while decreasing in year 2-5 is not a big deal, it becomes more of an issue as the cost of financing goes up, so the main effect is to increase the already-high sensitivity of software development to tightening credit.
No expensive machines or lab equipment, sometimes not even office space needed, just people with laptops
But the cycle wave length should ... Be similar to how many years people spend studying? That doesn't seem to be the case here?
Quoting the Wikipedia link:
"high salaries in a particular sector lead to an increased number of students studying the relevant subject; when these students enter the job market at the same time after several years of studying, their job prospects and salaries are much worse due to the new surplus of applicants."
Maybe you could say it's related to investors' pork cycle (of different/unknown wavelength) not students' pork cycle :-)
Or maybe just one-time opportunism: low interest rates etc
Like, forcing some not very edge case friendly set of text field validation rules onto the beurocrats such that he can't just do what he tries to do as he could with paper work.
In many orgs. nowadays the processes seem to be made to fit the computer programs not the needs of the beurocracy.
... to begin the loop again
I suspect its more a structural issue
e.g. because of Big VC (a16z, other firms with dozens or hundreds of investing staff), VCs don't stick around firms long enough for real returns (cash distributed) to matter. If you're at a place and your stuff is marked up 10x after 4 years, you just hop to become a GP and try to ride the next markup wave. Even if these people exit VC after 10 years that is still 10 years of deals that all flopped.
This might not even be the individual VC's fault -- there may just be too many venture dollars chasing too few power law returns, so you get a surge of startups circa 2019-2022 that all disappear
The problem is that in this non-ideal world value creation can become completely detached from returns: when instead of investing into people creating value, or distant future returns, or even medium term returns, it becomes about investing into other people investing in it—not too different from a pyramid scheme and various pump-and-dumps.
FRED is just a descriptive tool: https://www.nytimes.com/2024/12/06/business/economy/fred-eco...
Where is there any indication that FRED wants this?
That interest rates started to rise right when the tax changes to make it less advantageous happened? 1-2 punch.
Last week I see the exact job I applied to posted yet again. Ghost jobs really need to be punished. If you aren't actively talking to candidates within X days of posting or are reposting a job before Y days, there needs to be some disincenive to not do that.
the whole point is to obfuscate growth or decline to blur the ability of business intelligence units to make prediction about the company.
it's a scam for applicants, and a scam on the market.
I'm not convinced H1B issues are related, though I'm sure there are fake adverts for jobs that are unfillable to get H1Bs in, too. Unfortunately Musk, Ramaswamy, and DOGE are all about cheap workers.
They were somewhere between ok and not good. Felt like we got about what we paid for - their cost was about 50% of a US dev. They were as productive as a low end US dev, so in some cases ok value but since hiring low performers tends to hurt a team overall they weren’t worth it.
We’re a mid range company (decent cash comp, not much else since startup). I’m sure if you’re paying more you can do better. But that’s how it’s always been hasn’t it?
Also really top devs (at least in Brazil) can make more than 50k USD per year (fluctuates widely based on exchange rates) at Brazilian companies, it is cheaper than US but the tail-end of talent can still be expensive.
If you really want to attract good devs in cheaper areas the best way is to just open a branch of your office there, pay top-salary and don't just use it a dump ground for the projects no one at home wants to take. So treat them the same as at home. I had a roommate who worked in one of those companies that take offshoring projects, lets just say it is not the best talent pool and the good devs there leave fast.
This is the usual ebb and flow of offshoring in the tech industry that's been going on since the 90s.
There has never been a scalable arbitrage for tech talent by going over to another geo, and there won't be one in the foreseeable future. What US HCoL talent gets paid is unfortunately the fair and natural market rate for that caliber of talent, at least if you need to hire more than like 10 people.
I’ve been hearing about outsourcing destroying jobs since the 90s and that’s how it’s always gone: the suits salivated at the prospect of cutting wages by, say, 90% and had total write-offs because of it, because they missed that even in a poor country people have options and anyone smart enough to be a good developer is also smart enough to recognize that their skills are worth more. Outsourcing has some big costs related to communications so while you could find decent people at 50-70% of local labor rates, coordination overhead makes that a net loss even before you hit things like the security risks.
Part of the problem there is that the business people really want to think that they understand their business so well that they can give perfect instructions, and it takes a certain humility to recognize that more time goes into knowledge transfer and discovering the true needs than might be obvious.
There are great engineers around the world but why would they accept ten beans an hour if they can take a plane to a different country and earn fifty?
The talent from LATAM are not just sitting in the rainforest waiting for your phonecall.
I reckon there's a lot of "garbage in garbage out" going on, and if an org took the time and effort to actually treat Indian (or any outsourced team) devs like their own, things will be much better. But when you do that the overhead shoots up, and that cost/benefit analysis you did during the proposal to offshore flies out the window.
It's happening, and happening fast.
The boring reality is that the price of labor roughly reflects the productivity companies typically get. Because if there is ever some clear win in some location, people start hiring there, and things start to balance out again.
Outsourcing is a cost savings measure. It's a symptom, not the cause of the issue
The demand for engineers during the pandemic may have been artificial, but there is still need for people today. Some companies don't hire because they were built to be sold.
I can't agree with the conclusion. I can come with tens of examples of when using AI doesn't work.
Also, software engineering means much more than writing code.
Now in the AI boom era, 90% of the capital goes into the AI while 10% goes into SE jobs. That could explain the drop in SE jobs.
What if it speeds up the part of pumping out widgets in the widget factory, shipping Spring Boot / Laravel / whatever projects more quickly and therefore individual devs becoming more productive on cookie cutter projects (without necessarily getting more money for it either).
The headline doesn’t tell the whole story. SE openings precipitated in the year that followed the peak. Compared to that massive decline, openings over the last year have been relatively stable but on a downward trajectory.
Most of us are personally familiar with the reasons for this. Companies overhired during COVID and the demand for engineers was met by higher CS enrolment and to smaller extent, boot camps. But then remote work made companies realize that they could hire all over the world. Right now Netflix is only hiring engineers in Poland and other Faang companies are mostly interested in South America.
Focus on AI shifts investment in technology into that field, regardless of whether it makes sense or not. Likewise management expects that AI will drive up engineer productivity.
There has been no significant decline in software engineering positions since the advent of AI as all the factors I mentioned above still apply in the post 2023 period.
This is patently false. Source: I’m an engineering manager at Netflix who’s hiring in the US and whose peers are all doing the same.
When was the peak of AI? My impression was around 2022, and the crash started the next year.
I don't think AI is the cause so much as projected economic headwinds and S174. But I can see AI being indirectly responsible. Not because of replacing jobs, but because Companies want to redirect dev funds saved into powering AI servers. I believe Meta said as much at the beginning of 2025 (or maybe it was Google?)
Just think how much the investor narrative has shifted from early 2020s. Back then everyone talked about the business areas they were expanding into using tech. Now, investors get hyped up when you do the same thing, at lower cost. Less about growth, more about predictability. Which makes sense. Do investors view Amazon as a high tech company (as in 2020 and before) or a predictable business? dare I say utility?
It explains layoffs, forced RTO, etc. They simply don’t care if even their best leave. Their MBA executives know optimization/keeping the lights on - and less about innovation.
So it's a bit of both extremes at once.
Few, maybe 15%, of those people really seemed like they could program. Everybody else talked about their favorite tech stack and their opinions on code styles. It kind of felt like code masturbation. A shrinking job market is not surprising.
It certainly feels this way if your only perspective on the tech industry comes from headlines and social media.
Step outside of the doomerism whirlwind that is internet headlines and it's a very different story. There are a lot of companies doing great things and delivering products that people use, but they don't make for catchy headlines. So you don't hear about them.
This also means it is a great time to start a small, profitable business because labor is undervalued.
Later moved to Oracle / SQL Server / C# / microservices / CICD etc etc) but hell, that was my whole career 2002-2012 across various companies. Making cool tools for smart people, things the accelerated the company in smal;l but measurable ways, that unbeknownst to me would absolutely have directly resulted in a team of 3-5 being reduced to a team of 1, or a team of 1 not being expanded to be a team of 3-5.
The same will happen (IS happening) with LLM's.
I’m not sure that’s the case even if in retrospect we can clearly argue this
In 1998, Paul Krugman, winner of the Nobel memorial prize in economic sciences, infamously predicted that “the growth of the Internet will slow drastically, as the flaw in ‘Metcalfe’s law’—which states that the number of potential connections in a network is proportional to the square of the number of participants—becomes apparent: most people have nothing to say to each other! By 2005 or so, it will become clear that the Internet’s impact on the economy has been no greater than the fax machine’s” [1] and even though that turned out to be spectacularly wrong it shows the attitude in the early days wasn’t one of absolute certainty.
And certainly the current AI boom is most visibly known for LLMs but there is a lot more happening beyond chatbots.
[1] https://www.snopes.com/fact-check/paul-krugman-internets-eff...
Even if everything being done with LLMs is useless, if you can do the same amount of useless with half as many developers, it's bad news for us.
The final phase of this is where Americans start moving to some hut in the Congo to survive off the market wages offered when the supply side of the supply/demand curve blows up.
Edit: I am an idiot, there is a mention. Or possibly the article was edited and I'm not an idiot, it does say there was an edit recently.
I originally didn’t mention this because S174 impacts the US and US-HQ’d companies. In this data other countries like Germany, UK, France all see a similar drop. Also, S174 impact likely really started from early 2024, when companies impacted had to pay high taxes and realized the change is here to stay with no end in sight. Doesn’t explain the drop since 2022.
Updated the article to make this clear though. It was not in the original version - thanks for the note!
Bullshit. Spotting hallucinations requires expertise in the subject matter, and most devs wouldn't be using LLMs if they intimately knew the programming languages and APIs the LLM is generating.
The reality is that spotting hallucinations often takes more effort than reading the source documentation and writing the code from scratch, since the dev also needs to review and check the code for correctness.
With Cursor and Copilot, the days of $190k entry-level SWE roles in the US are GONE.
When it does save me time, it's almost just some boilerplate I didn't have to Google or type out. Honestly, I feel like gen AI pleataued a year or more ago.
LLMs are just really convincing bullshit generators. They look impressive on the surface, and the times when they spit out a whole bunch of useful boilerplate feels like magic, but that stuff isn't super useful for a majority of the work you spend your time doing. Throwing more money at these companies is not magically going to yield AGI. I think the AI CEOs are basicallly selling us a lie.
I work in typescript, rust, and go - mostly typescript. LLM coding is an order of magnitude better at typescript than these other languages. To contrast, LLMs seem mostly useless for rust.
I also use cursor, which is a big jump over other code assistants.
And finally I understand how to prompt LLMs accurately. This is a new tool and from interacting with coworkers in the same codebase I can say many smart people have not yet learned how to use the tools effectively.
But if you just assume that everyone catches up to where I am today with cursor + typescript the change is massive.
It takes a bit time to build intuition around the workflow, but when you get going, it seems surprisingly useful. I was a skeptic before as well, btw.
Juniors need to learn by falling on their face a few times, if companies don't want to train them then they'll never become seniors. Lovely cycle
It's rare that anyone gets trained by a company who doesn't already know something and is pivoting to a new role. It's unfortunate but that's just the way it's always been. Juniors have to be proactive and bust their ass to become mid level, usually by learning new stuff at home through building stuff on their own and reading obsessively. The hardest part for a junior to get is their first paid job, and more often than not, that first job will be at a shit hole. Juniors should take advantage of it and really bust their ass to impress them and improve themselves. Once they get a couple of strong learning years under their belts, it gets a bit easier.
If you are a junior and the company is paying you to learn stuff, consider yourself extremely lucky.
Companies are in desperate need of strong talent. To succeed, juniors need to force themselves to become strong talent.
Some hallucinations stay in codebases longer than others! If there were zero hallucinations there would really be no novel output. Some hallucinations are useful and some are not.
Facts don't really play a part there, if a response is factual its only a sign that the training set largely agreed on the facts (meaning the correlation of token sequence was high).
Also, is this the downside of the 'wait and see'
But, there could be a rebound from this.
" Software by non-developers creates more opportunities for devs. Imagine a situation where the number of non-developers creating software increases by 10x or 100x, due to AI, thanks to non-technical people creating software with AI tools and agents. "
Just like with arts/video editing.
There could be a rebound when so many people start creating things with AI,
THEN realize it is actually a bit harder than they thought, and boom, need to hire video or software engineers to finish what they started.
- Offshoring, nearshoring
- Generative AI
- Coding camps
- Better and more open source libraries. More high level open source libraries
- Low code and no code
- Easier to use technology, more documentation
Smaller teams and more bootstrapping (as opposed to venture-funded rapid growth) seems likely to reduce the reliance on recruiters who are a plague on the industry, with few exceptions.
On the other hand, maybe a lot of tasks you could previously get a lot of leverage on with a simple Perl script will just be done directly by LLMs. Not if they're customer-facing, maybe, but in cases where you just need to get some data in a different format or something.
Assuming that AI coding is already helping the average dev with shipping more.
A better graph would be of the number of people who self-report as being employed as software developer. I doubt that would show the same decline--there have not been many stories of people abandoning the profession because they could not find employment...
In 2000, none of the ideas ended up being bad. They were just too early and internet wasn’t ubiquitous.
But even then, if you were working as a standard enterprise dev for a profitable business - banks, insurance, companies, etc, outside of Silicon Valley, jobs were plentiful. I was a Windows dev living in Atlanta with 4 years of experience and had no issue of getting offers.
Things improved when internet became ubiquitous both at home and in their pockets with smart phones and the App Store by 2012.
On the B2B side, there was the rise of SaaS.
There will be no next doubling of jobs to absorb all of the people looking. Every application gets literally 1000 applications within the first day of posting.
It’s going to take at least a decade for the supply/demand to balance out.
No Section 174 is not to blame either.
If H1Bs were so destructive to the software industry, why would the software industry, which has the highest prevalence of H1Bs, be the industry with the highest growth in salaries for basically the past 2 decades?
Also, the idea that eliminating H1Bs will bring more jobs makes absolutely no sense.
If a company is asked to eliminate all their H1B positions, what do you think is more likely:
- they hire a completely new employee in the U.S. at a similar, or if the critics are right and H1B is suppressing salaries, higher salary, or
- they hire the exact same person but now with that person back in their home country and pay them a tiny fraction of what they were paying them in the U.S.?
There may be a handful of jobs where the employee needs to be in the U.S. where sure, a few more Americans will get jobs. But the overwhelming jobs will simply move abroad with the H1 employee, and along with that a whole bunch of dollars that were being spent within the U.S. and was paying US taxes will move abroad and be spent abroad and pay taxes abroad.
Maybe because that’s just not true? Salaries have generally been stagnant over the last 20 years in real terms for most in the industry.
Most developers do not work at FAANG positions, and even those positions are only found in the extremely high cost of living areas like Silicon Valley.
Companies are always free to hire people from wherever they want. However, it could be foreseeable that “American” companies that are largely overseas, maybe at a disadvantage when it comes to taxes or other benefits conferred by the US government. Such companies may find that their products are subject to tariffs when trying to sell those products back in the United States or other active measures to prevent offshore positions. There is a risk as section 147 changes already has shown.
I thought it was the MySpace of job boards.
Today: "I know this pain! It's real."
Rough out there.
this is a very stupid thing to say. The formal structure of a language doesn't easily determine the complexity of what can be done with a language, nor is it a way of predicting the difficulty of expressing an equivalent idea vs other formal or natural languages.
Well, it does have the virtue of being straightforwardly true. For example, programming languages can't be ambiguous.
This is "straightforwardly" false.
Speaking specifically about the "language" part - while there is a formal specification for C++, many pieces are implementation dependent. Then there is actual undefined behavior that is part of a specification.
Actual programs implemented in a language can be ambiguous. consider multi-threaded programs where data arrive at different times in different threads, leading to different outcomes. Or just pure ambiguity of intent. Or a program which incorporates undefined behavior intentionally.
Formal and natural languages may overlap in some ways but it is ridiculous to compare them in this way and claim a probabilistic model is better at the formal language. Translation tasks are an example where LLMs perform extremely well, I would argue much better than in programming. Should I make the claim it's because of some intrinsic attribute of natural language vs formal language?
Yes, in C++ we call it "unspecified behavior" and "undefined behavior"
And what happens if you do?
