The one about the web developer job market
We have the worst job environment for tech in over two decades and that’s with the “AI” bubble in full force. If that bubble pops hard before the job market recovers, the repercussions to the tech industry will likely eclipse the dot-com crash.
Executive summary: #
- The developer job market is unlikely to ever fully bounce back.
- Tech companies are still performing mass lay-offs, even with a tech bubble in full force.
- Many organisations are also resorting to employee-hostile strategies to increase employee churn, such as forced Return-To-Office policies.
- Web media has historically been a major employer of web developers and designers, but the entire sector is threatened by the simultaneous decline of search engine quality and search-replacement tools that do not direct traffic to their sources.
- Large Language Models (LLMs) are unlikely to lead to a wholesale revolution in software quality. (Productivity, Return-On-Investment, and Quality are not synonyms. LLMs only have a plausible case for increasing one of those variables.)
Developer summary #
- Finding a non-bullshit job is likely only going to get harder.
- The overall developer job market will continue to fluctuate, but without dramatic change there isn’t much on the horizon that seems likely to turn the decline around.
- Finding effective documentation, information, and training is likely to get harder, especially in specialised topics where LLMs are even less effective than normal.
Predictions are a mug’s game, but occasionally you just have to be a mug #
Tech predictions, specifically, are a mug’s game. Trying to accurately estimate the impact of a single variable, such as a technical innovation, on a complex system is futile, even if you manage to isolate said system.
When you have multiple complex systems that are intertwined – social, economic, cultural, political – predictions become especially foolish.
Forecasts can be useful: short term estimates of the likely trajectory of single variables are both possible – as long as you pay attention to your margin of error – and useful. But as soon as you start to try to predict the second or third order consequences things very quickly get remarkably difficult.
With extremely complex modelling, meteorologists can estimate where a newly-formed storm-front could land. They won’t be able to pinpoint exactly, but can give you a likely range.
Predicting the exact consequences of that landfall is much harder, and predicting the consequences of those consequences is impossible.
In short, futurists are largely con artists.
Accurately predicting the web developer job market is effectively impossible. It exists at the intersection of multiple different economies, social systems, and cultures. It’s governed by multiple different interconnected variables with interlinked consequences.
But when you’re faced with a potential disaster, such as the potential eradication of your entire profession, you need to do something to get a sense of the broad range of consequences you might be facing.
Predictions are a big part of planning for natural disasters and both geological and weather systems are easily some of the more complex systems you can encounter today.
What sort of predictions can you make for a complex system?
When your region has a volcano habit #
For the past five months, the Svartsengi region in Iceland has been suffering from regular geological instability. Every month – November, December, January, February, and now March – we’ve had major magma activity, usually followed by an eruption.
Making exact long-term predictions about this geological system is next to impossible. The activity could continue for weeks, months, years, or decades. We don’t know. The eruptions could all be the same size, or they could escalate. They could erupt in areas that are far away from infrastructure and populated areas, or magma could spew out of the ground right in the town centre of Grindavík. We don’t know how much toxic gas each eruption with spew out, nor do we know where that gas will go afterwards.
But you need to save what can be saved and protect people’s homes and lives, and for that you need estimates. You can estimate the likelihood of harm happening and then take steps to make that harm less likely.
- Could lava flow over the town or the power plant? Yes.
- Is it possible to make the lava less likely to destroy the town or other infrastructure? Again, yes.
- Do you need to know exactly what will happen – where and when – to build that protective infrastructure? No, you don’t.
So we build massive earth walls and hope for the best. Sometimes they work and divert the lava flow away from the town and power plant. Sometimes they don’t hold. Sometimes the eruption happens inside the barrier, bypassing it completely.
Some harm was prevented. Not all of it. But we’re better off for trying.
We can try to apply the same principles to figure out where we stand, as workers or employers, in the larger web developer job market.
We can figure out the broad range of possibilities: worst and best case scenarios.
We can then estimate what measures we can take to minimise the impact of potential disaster on our own lives and work.
Some harm will be prevented. Not nearly all of it. But we’ll be better off for trying.
The state of the tech job market today #
With tech layoffs at their highest since the 2001 dot-com crash, the job hunt is getting harder and many in the industry are being forced to settle for pay cuts if they can find a new gig at all.
Laid-off techies face ‘sense of impending doom’ with job cuts at highest since dot-com crash
The last time employment in tech was this bad, in the early 2000s, I dealt with it by simply opting out. “Fuck it, I’m doing a PhD”, and I just didn’t participate in the developer job market for four years.
Well, I lie. I still did freelance gigs on the side, but because I managed to get a grant for my PhD work, a few months passing between paid projects wasn’t a disaster.
I’m not about to do another PhD – one’s enough.
The software developer job market today looks at least as bad as it did in the aftermath of the dot-com collapse.
This is not a one-off event but has turned into a stock market driven movement towards reducing the overall headcount of the tech industry.
Two words explain this year’s trend: stock prices.
And this trend towards job cuts seems concentrated in tech or tech-adjacent sectors such as streaming.
The Technology sector leads all industries this year with 28,218 cuts, 12,412 of which occurred in February.
Job Cuts Jump in February 2024; YTD Cuts Down 8% Over Last Year
This is despite the arrival of a boom in funding due to the generative model bubble. “AI” is ostensibly here to turn the tech industry around.
Every month sees massive investments in “AI” and related tech:
In total, the first three weeks of the month have already seen more than $2.6 billion invested — including a half-dozen rounds of $100 million or more — with still a week-and-a-half left.
Tech bubbles are not a recent phenomenon. The dot-com bubble was far from the first of its kind. But this is, as far as I can tell, the first time when we simultaneously have a bubble-proportionate increase in tech funding and investment and a post-bubble sized downturn in the tech job market.
Those two are supposed to be successive – first one, then the other – not simultaneous.
What this means is that when the bubble ends, as all bubbles must, the job market is likely to collapse even further.
It’s impossible to predict the duration of a bubble. Some are short. Others last for years. But they all end eventually, and they all leave with the job market worse than it was before.
The reasons for the disconnect between the tech industry’s overall prospects and the job market seem twofold:
1. The stock market loves job cuts #
This is even though all research and evidence points to them being both expensive and disruptive to a business’s ability to generate profits.
2. Activist investors see it as an opportunity to lower developer compensation #
From the horse’s mouth, Christopher Hohn of TCI Fund Management Limited wrote the following in a letter to Sundar Pichai, CEO of Alphabet (née Google):
I believe that management should aim to reduce headcount to around 150,000, which is in line with Alphabet’s headcount at the end of 2021. This would require a total headcount reduction in the order of 20%.
Importantly, management should also take the opportunity to address excessive employee compensation.
3. Management believes they can replace most of these employees with LLM-based automation #
No, really.
Anyone else hearing this? My boss, the CTO, keeps talking to me in private about how LLMs mean we won’t need as many coders anymore who just focus on implementation and will have 1 or 2 big thinker type developers who can generate the project quickly with LLMs.
Executive leadership believes LLMs will replace “coder” type developers (Reddit)
Speaking at the Word Government Summit in Dubai, Huang argued that because the rapid advancements made by AI, learning to code should no longer be a priority of those looking to enter the tech sector.
I would argue that the vast majority of Classical CS becomes irrelevant when our focus turns to teaching intelligent machines rather than directly programming them. Programming, in the conventional sense, will in fact be dead.
“This is a much bigger disruption than the pandemic,” he told the audience, going back to his point about AI large language models successfully writing software code. OpenAI’s ChatGPT can pass Google’s exam for a high-level software engineer, Mostaque said, even though it’s a non-specialized model. “There’s no programmers in five years,” he predicted.
Stability AI CEO says AI will prove more disruptive than the pandemic
The fact that pretty much everybody quoted to say that programmers will be replaced with LLMs is either a CEO or CTO is important.
These are the people deciding to do lay-offs.
Whether it’s actually true or not doesn’t change that this is part of what’s driving the decision to cut jobs en masse.
Discovering whether it’s true or not is actually quite complicated as it, counter-intuitively, doesn’t depend on the degree of LLM functionality but instead depends entirely on what organisations, managers, and executives are using software projects for.
The role of LLM-based coding tools in the developer job market #
Not every CEO is predicting the end of programming as a profession. GitHub’s CEO instead thinks that programmers will be the primary beneficiaries of the introduction and improvement of LLM tools for coding.
With Copilot, if you only have two to four hours a day to actually code, you can use that time better. You can use that time to stay in the flow, to get the job done and enjoy doing it.
GitHub CEO says Copilot will write 80% of code “sooner than later”
This difference of opinion isn’t just down to the fact that GitHub’s customers are developers and would suffer if they were eradicated as a profession. (Although…)
The difference is a question of different attitudes towards software development in general and varying opinions about how software generates value.
Productivity versus churn #
The positive case for LLMs is to claim that they will remove the drudge work of coding.
That isn’t as big of a productivity improvement you might think, as GitHub’s CEO Thomas Domke explains:
Developers don’t actually spend most of their time coding these days — between two and four hours per day is when a developer writes code. The rest of the day they do other things, like stand-up meetings, looking at crash reports.
GitHub CEO says Copilot will write 80% of code “sooner than later”
Since, even in the best case scenario of the most optimistic prediction of LLM power, you’re still going to need to structure a plan for the code and review it, the time spent on code won’t drop to zero. But if you believe in the best-case prediction, a 20-40% improvement in long term productivity sounds reasonable, if a bit conservative.
This world-view assumes that the purpose of software development is the productive creation of successful, defect-free, software projects. LLMs would increase productivity.
The alternate world-view, one that I think is much more common among modern management, is that the purpose of software development is churn.
- Faster feature development to drive sales to customers who then are locked into your software.
- Presenting an image of an organisation that is working with the latest technology.
- Responding quickly to the whims of investors and the stock market.
- Bolstering a manager’s resume or status within the organisation.
None of these require that the software be free or even low on defects. The project doesn’t need to be accessible or even functional for a majority of the users. It just needs to look good when managers, buyers, and sales people poke at it.
The alternate world-view, which I think I can demonstrate is dominant in web development at least, is that software quality does not matter. Production not productivity is what counts. Up until now, the only way to get production and churn has been to focus on short-term developer experience, often at the expense of the long term health of the project, but the innovation of LLMs is that now you can get more churn, more production, with fewer developers.
This means that it doesn’t matter who is correct or not in their estimate of how well these tools work.
A note on the range of improvement from LLM-based coding tools #
The range of improvement people propose for LLM coding tools is vast.
On the low end you have people like me, who think it’s going to be a net detriment to the field, on the high end you have people Thomas Dohmke, the GitHub CEO quoted above, who think it’ll supercharge programming as a profession.
Some of that difference is down to bias. My entire shtick is that the software industry as a whole is self-evidently dysfunctional because most software is either broken in some way or doesn’t even end up shipping to users (because it was too broken). It shouldn’t be surprising that somebody who thinks the industry is incapable of shipping a well-made software product is of the opinion that the industry is incapable of actually shipping a safe and functional LLM-based software product. GitHub’s CEO, conversely, is in the business of selling said LLM-based software product. He’s clearly not likely to focus on the flaws of his product.
But much of it also comes down to perspective. The most dangerous flaws in LLM coding tools are sporadic.
-
Training data poisoning is impossible to prevent, hard to detect by the end user, and potentially devastating. You aren’t going to notice the issue as an end-user. From your perspective the system is working perfectly.
-
Security issues in code contributions in general are hard to detect, especially if the language and platform is one you don’t have senior experience with. LLMs are prone to generating unsafe code. Experienced developers will edit out the issues without thinking about it, focusing on the time-saving benefit of generating the rest. Inexperienced developers won’t notice the issues and think they’ve just saved a lot of time, not realising they’ve left a ticking time bomb in the code base.
-
Code copying is really hard to detect because LLMs tend to paraphrase output. Even verbatim copying happens about 1% of the time according to GitHub’s own (biased) research. Copying code and just changing the variable names wouldn’t be detected by their safeguards but is just as much a copyright and licence violation as verbatim-copied code.
-
The training data favours specific languages such as Python and JavaScript. Even within JS, it favours Node and React. This is purely by virtue of the amount of relevant code in the training data sets. Developers who favour these platforms will get a marginally better experience with an LLM than somebody who use a less well represented language or framework.
-
Beyond the ability to spot flaws in LLM output, experienced developers are also much better at using LLMs in ways that enhance productivity without exposing their code base to as much risk. The same tool that enhances their productivity by 20-30% might also be outright harmful to a junior developer’s productivity, once you take the inevitable and eventual corrections and fixes into account.
It’s hard to gauge the harmful effects of LLMs on the field as a whole from the perspective of a single developer testing the tools directly. The risks only become obvious once you look at studies that look at the effects across a sample size that’s larger than a single person, and from that perspective LLMs look like a Y2K-style code quality disaster in the making.
But that’s not relevant to the argument in this piece. From the job market perspective, all that improved and safe LLM-based coding tools would mean is more job losses.
Because the effect LLMs have on the job market is down to manager world-view not the level or quality of the innovation.
Why manager world-views are more important than LLM innovation #
As I noted above, I’m on the record as believing that these systems aren’t nearly as functional as claimed and that their rate of progress is much slower than promised.
GitHub’s CEO believes the opposite.
It doesn’t matter which one of us is right. If the technology is what’s promised, the churn world-view managers will just get more production, more churn, with even fewer developers. The job market for developers will decline.
If I’m right, they will still use the tech to increase production, with fewer developers, because software quality and software project success isn’t what they’re looking for in software development. The job market for developers will decline.
When most of the tech industry’s management class is intent on removing you from the equation, successful innovation and progress in automation is inevitably bad news. The more progress you see in the automation, the fewer of us they’ll need. It isn’t a question of the nature of the improvement, but of the attitudes of management.
Even if that weren’t true, technical innovations in programming generally don’t improve project or business outcomes.
Take automatic memory management as an example. It’s a genuine improvement to programming. All else being equal, automatic memory management of one kind or another will almost always improve programmer productivity, code safety, and can even improve the performance of the final software.
But it doesn’t make a software project more likely to succeed in terms of business outcomes because that’s governed by how well the software works for its end-users. The odds of a project’s success are dictated by user research, design, process, and strategy, not the individual technological innovations in programming. Rapid-Application-Development tools, for example, didn’t shift outcomes in meaningful ways.
That’s the spirit behind Fred Brooks’ No Silver Bullet paper. Programmers throughout the years who try to debunk the paper by tabulating the results from various technological innovations are missing the point. What matters is whether the final product works and improves the business it was made for. Business value isn’t solely a function of code defects. Technical improvements that address code defects are necessary, but not sufficient.
When management of a project is disconnected from business value all that technological innovation in software development will do is change the flavour of dysfunction.
It requires that management believe in the value presented by software developers and designers.
And as you can see from the quotes above, tech industry management is firmly convinced that less is more when it comes to employing either.
Whether you’re a bear or a bull on LLMs, we as developers are going to get screwed either way, especially if we’re web developers.
Because web software quality, more than any other sector in software development, is disconnected from business outcomes.
Most web projects shipped by businesses today are broken, but businesses rarely seem to care.
Most websites are buggy, inaccessible, and broken #
The overall dysfunction of the web software industry has been demonstrated over and over by a number of people:
- The Website Obesity Crisis
- The Performance Inequality Gap
- JavaScript Bloat in 2024
- How web bloat impacts users with slow connections
Those are just the posts that I’ve come across in the past week on social media. Most websites are inaccessible, even in countries where that exposes the business to substantial legal liability. Most websites perform so badly that they don’t even finish loading on low-end devices, even when business outcomes directly correlate with website performance, such as in ecommerce or ad-supported web media.
A large part of web development is disconnected from business success and software quality.
The current state of web development is as if most Windows apps released every year simply failed to launch on 20-40% of all supported Windows installs.
It’s clear from even just a cursory look at the data presented by the authors of the posts I listed above that web project quality and functionality isn’t what matters to many organisations. Instead, what matters is being seen making projects that have a plausible semblance of functionality and technological progress because that’s what drives investment and stock prices. If being plausible is all that matters, then that’s the literal, genuine, core strength of an LLM.
The disconnect between project quality and business success happens for a variety of reasons.
- High software margins (you still have a profit despite the dysfunction)
- Organisational tolerance for partial success
- Fresh codebase momentum leading to customer lock-in
- Managerial career chess
- Pure vanity theatrics
You don’t need to look far on the web to realise that web projects, for the most part, don’t need to work for general users to be seen as acceptable by their owning organisations.
This is a problem for the job market because if all that matters to these organisations is being seen plausibly chasing cutting-edge technology – that the actual business outcomes don’t matter – then the magic of LLMs mean that you don’t actually need that many developers to do that for much, much less money.
Judging by the sheer volume of utterly broken websites on the web, these organisations represent a non-trivial portion of the current web developer job market.
This might be manageable on its own if it weren’t for the fact that we’re on the verge of losing an entire sector of the web industry
Web media: the eradication of an entire sector of employment #
If you haven’t been paying attention to the web media scene, it’s been having a hard time. A number of companies have been having trouble, laying off much of their staff. Not because they, like tech companies, believe that the workers can reliably be replaced by generative models. They’ll try, but those experiments quickly end up being disasters. But the core reason for the lay-offs is that much of web media is already in free-fall.
As Josh Marshall, of Talking Points Memo fame, wrote:
I’m reposting the chart here just for the purposes of illustrating what I’m discussing.
As I made clear in the post, the chart tells part of the story of one company’s successful efforts to grapple with the vast changes in the programmatic and larger advertising landscape over the last decade. On an apples-to-apples basis, revenue from programmatic advertising hasn’t gone down by 95%. This steep drop off is also due to changes we made in reaction to the cratering of the programmatic ad market.
Josh Replies to the Slings and Arrows of Those Claiming Digital News Advertising Is Going Great!
Subscriptions and native advertising aren’t coming close to making up for a 95% shortfall.
As he continues to say in the above linked piece:
The first of which is, who are we trying to kid here? Does anyone think that advertising — direct or programmatic — still sustains digital news organizations, especially independent ones? Really? I think the almost weekly lists of bankrupt and shuttered news outlets tells the story pretty clearly.
Web media is a major employer, both directly and indirectly, of web developers. If a big part of the web media industry is collapsing, then that’s an entire sector that isn’t hiring any of the developers laid off by Google, Microsoft, or the rest. And the people they aren’t hiring will still be on the job market competing with everybody else who wouldn’t have even applied to work in web media.
This would be bad on its own, if it weren’t for the fact that search engine traffic is declining as well. LLM-enabled spam sites are flooding the search engine results which drives down traffic to web media sites in general.
The bad news is that it doesn’t seem like this will get better any time soon. The study points out generative AI sites one or two times, but that was only in the past year. The elephant in the room is that generative AI is starting to be able to completely automate the processes of SEO spam. Some AI content farms can scan a human-written site, use it for “training data,” rewrite it slightly, and then stave off the actual humans with more aggressive SEO tactics.
And it’s unlikely that Google will be able to adequately counter this growth in SEO spam. As Mark Williams-Cook of Search Engine Land says:
I believe this to be the current situation:
- Google’s current ranking systems can’t keep pace with AI-generated content creation and publication.
- As gen-AI systems produce grammatically correct and mostly “sensible” content, they pass Google’s “sniff tests” and will rank until further analysis is complete.
Herein lies the problem: the speed at which this content is being created with generative AI means there is an unending queue of sites waiting for Google’s initial evaluation.
The scale of LLM-enabled spam production outstrips the ability of Bing or Google to counter it. Results filled with spam pages lower the traffic directed by search engines to web media. Lower traffic means less native advertising and subscription revenue, which were supposed to offset the 95% drop in programmed advertising.
But it gets even worse as every major search engine provider on the market is all-in on replacing regular keyword search with chatbots and LLM-generated summaries that don’t drive any traffic at all to their sources.
As Neil Patel writes:
Search Generative Experiences (SGE) and ChatGPT Usage Will Drive Down Traffic from Organic Search.
These changes will inevitably decrease organic search traffic from Google because the SGE feature will siphon off some greater than zero percentage of searches. In other words, people will find enough of what they need in the SGE and not click on organic results.
Why send traffic to web media sites when you can keep it on site and take all the ad revenue for yourself?
Between the push to replace search with chatbots and the search engines themselves filling up with spam, it’s inevitable that the web media sector will shrink, possibly even disappear. Those that survive will have to operate on a tighter budget and have less scope for investing in web development.
That’s an entire industry removed from the web developer job market.
Where can things go from here? #
It’s reasonable to expect that the job market is unlikely to ever fully bounce back, due to the collapse of web media alone.
It’s also reasonable to expect that the job market might take another sharp turn to the worse because the AI Bubble will run its course eventually. It doesn’t matter whether it’s a genuine innovation or an overblown yarn-ball of dysfunction and wishful thinking, bubbles end eventually.
Both finding a job and hiring for web development will likely only get harder.
Due to the existing dysfunctions in the web development scene, there are very few signals you can rely upon to easily filter job applicants. Experience in Node or React is not a reliable signifier of an ability to work on successful Node or React projects because most Node or React projects aren’t even close to being successful from a business perspective. Lack of experience in Node or React – such as a background in other frameworks or vanilla JS – conversely isn’t a reliable signifier that the developer won’t be a successful hire. Pretty much every measure used by companies in hiring web developers is full of noise, and you don’t solve that problem by adding more noise with an LLM-based tool.
This means that one of the risks we’re facing in the industry is information asymmetry in the web developer job market could turn it into a market for lemons.
This is not inevitable and the consequences are dire enough that it would benefit employers if they took measures to minimise the odds of it happening, such as increasing their transparency, cooperating with labour unions and unionisation efforts, and resisting the urge to lower pay.
The potential “lemon” dynamic we are facing is that employers both have a substantial number of unfilled positions and the job market has a vast surplus of available talent, but because of the information asymmetry – employers can’t tell if somebody’s experience is valid for their projects, job-seekers can’t tell if the employer is one that genuinely needs effective web development – the market overall could decline, both in terms of the effort required to find a job and the pay you get once you’ve found it.
Lower pay, combined with the information asymmetry about employer dysfunction, would then lead to more capable workers leaving the sector, either to run their own businesses – a generally dysfunctional web development sector is likely to have open market opportunities – or leave the industry altogether. This would exacerbate the job market’s dysfunctions even further, deepening the cycle.
We don’t know how much of this will be offset, if at all, by under-served demand for developers from other sectors that have previously found it hard to compete with the tech industry for talent.
We don’t know if tech companies will change their outlook if they have some sort of software quality crisis. Since a crisis of that kind is likely to coincide with a downturn in tech, it’s unlikely that it would be a net benefit to the job market.
What we do know is that the job market is in poor shape today and that it’s likely to get worse before it gets better.
How much worse and how much better is impossible to predict – they’re exactly the kind of predictions I warned against above at the beginning of this piece.
The only thing we can do is assess our own personal risk and situation and act accordingly.
So, what should I do? #
Here we come to the reason why I did this research and wrote this analysis. I keep asking myself: what should I do?
I’m a freelance web developer with an ebook business on the side. I’ve been making websites since the late nineties and have managed to outlast several bubbles. But normally the bubbles themselves came with decent job markets. Our current situation where we have simultaneously a funding bubble and a poor job market is, in my experience, unique, certainly unique enough for me to try to look closer at what’s happening.
Since it’s impossible to predict specific shifts in the job market, our starting point always has to be to look at ourselves. “What can I do?”
Unionise #
Historically, whenever management adopts an adversarial attitude towards labour, the only recourse labour has is to unionise. Collective action limits the abuses of management and unionised industries tend to be more stable. This might have been a detriment to the sharply growing tech industry of a decade ago (if you believe anti-union messaging, that is), but today we are facing being outright cut out of the success of the industry. Why should we care if unionising hypothetically dampens the overall growth of the industry? It definitely dampens the downturns and since we are being cut out of the upside anyway, we might as well take measures to increase our job security.
As employees, we have nothing to lose from unionising. That’s the first consequence of management deciding that labour is disposable.
When you’re disposable, you might as well unionise. Thankfully, there’s a book for that.
This does not help people like me, though. I’m a freelancer and I live and work in a country that is 90% unionised anyway.
Diversify your skills #
Popular frameworks and languages are likely to be harder hit than others.
Web development, dominated by Node and React, is objectively more dysfunctional than much of the rest of software development to begin with. Just look at the number of websites that outright don’t work for a substantial portion of their intended audience.
Popular languages and frameworks, again like Node and React, are also overrepresented in LLM training data sets. This means that the quality of their output, thus their ability to drive down wages and employment, is likely to be higher for React, Node, Java, and Python, than it is for newer, less well-represented languages such as Rust or Zig. Newer, more recent, methods and approaches to web development will also be less well-represented.
Diversifying your skills has always been a good idea for a software developer. Learning a new language gives you insight into the craft of programming that is applicable beyond that language specifically.
But in this case it might be necessary for survival.
The problem is that finding good training resources is likely to get harder.
The developer benefits the most from learning languages and platforms that LLMs don’t handle well. That also means that the chatbots people are popularly using today to teach themselves new programming skills are less likely to do a reliably good job explaining those languages.
But the market for developer training in general has collapsed.
As Chris Ferdinandi says:
Sometime last year, sales of my courses and workshops started to decline quite a bit.
I’ve chatted with other folks who teach developers, and most of them are seeing similar things in their businesses as well. It’s a trend across the board for nearly everyone.
Some of it is down to the job market. Why invest in training if tech cos aren’t hiring you anyway? Why invest in training your staff if you’re planning on replacing them with LLM tools anyway?
Some of it is down to people using chatbots for learning instead.
Either way, your options for diversifying your software development skills are likely to become more limited over the next few years as the training market collapses in lock-step with the job market.
In my case, I’m certainly rethinking my original plan for this year, which was to focus on creating ebooks and courses diving into several recent innovations in web development, such as buildless JS with import maps, new approaches to CSS using CSS Layers, or full-stack SQLite. I’ll probably finish one or two of those projects, primarily the ones where much of the work is already done, but unless something dramatic changes I don’t think I’ll be beginning any new projects along those lines.
I should probably spend that time having another look at Rust or Zig.
Some of us could take the opportunity to make products #
The pervasive dysfunction of the web industry, combined with the monomaniacal focus tech companies have on only making products that involve LLMs or diffusion models in some way, might translate into opportunities here and there in making effective products for sectors that are underserved by the software industry.
This path is risky but, at this point, so is just trying to build a regular career in software development. Bootstrapped entrepreneurship might begin to look like a promising alternative to many in the field.
After all, as Amy Hoy wrote in 2016:
Running a biz is a lot less risky than having a job, because 1000 customers is a lot less fragile than 1 employer.
None of this will be easy #
The problem with giving people bad news is that you don’t always have something positive to say that can outweigh the bad.
Sometimes all the news you have to tell is bad.
I don’t think it’s safe to assume that the software developer job market will rebound. I’m hoping it will stabilise at a sustainable level, but even that isn’t guaranteed. The tech industry has “innovated” itself into a crisis, but because the executives aren’t the ones out looking for jobs, they see the innovations as a success.
The rest of us might disagree, but our opinions don’t count for much.
We aren’t helpless. There are tactics we can apply, avenues we can explore, and constructive efforts we can put our energies into.
But what we can’t do is pretend things are fine.
Because they are not.