> Management believes they can replace most of these employees with LLM-based automation
I don't think this is materially true, it's just a nice narrative. (And a way to entice investors, which, remember, is the CEO's job.)
> Most websites are buggy, inaccessible, and broken
It's been like this since at least 2010, so imo this isn't really anything new. We're just buggy in different ways. Switched from one footgun (jQuery) to another (React).
> That’s an entire industry removed from the web developer job market.
I'm not exactly sure what the "web media" sector is. Re-read that section like twice and I'm still confused. Programmatic advertising is dipping because things like FB/Google are losing out to TikTok and Instagram (& more tailor-made advertising). As someone that used to work in ad-tech, I don't really think those jobs are going away, just evolving in a post-Cambridge Analytica world. This has been in the works long before the LLM craze.
> Here we come to the reason why I did this research and wrote this analysis. I keep asking myself: what should I do?
The answer is always the same: start a company. I'm not sure why so many developers are terrified by the prospect.
> Sometimes all the news you have to tell is bad.
I dunno, I don't really see the doomerism here. I tend to be more optimistic than most, but even so, even if you can't get a job as a "web developer" (which really hasn't been a job title for about a decade now anyway), pivoting into data engineering, backend work, LLM work, infrastructure, etc. is not really that big of a leap.
> The answer is always the same: start a company. I'm not sure why so many developers are terrified by the prospect.
Perhaps because starting a company is an extremely high-risk endeavor that requires a different skillset than development does, and demands a great deal more energy, time, sacrifice, commitment, and tolerance for loss than any software development position.
It's worth doing (and I recommend it for those who are up for it), but it's absolutely not for everybody.
> The answer is always the same: start a company. I'm not sure why so many developers are terrified by the prospect.
Every one says this yet almost all companies fail in their first year. Surely you see why people would feel more comfortable being employed by someone else and having the security of a regular paycheck.
> 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.
Would generally agree and it’s been discussed a lot but apart from “because they can” I haven’t heard a really compelling reason why. I’m not convinced by the claims about AI taking jobs (yet). So what’s going on?
Companies that are not launching successful products and don't have competition pushing for innovations on their current products are firing R&D people.
I'm really tempted to add an ironic "News at 11!" at the end, by the way. But in case you are one of the large majority that didn't notice, currently the software market has a problem. And we can expect more bad consequences of it.
> The fact that pretty much everybody quoted to say that programmers will be replaced with LLMs is either a CEO or CTO is important.
Out of those four quotes, we have one from nvidia, one from fixie.ai, and one from openai. That may be important too. Either because these people are in positions to see trends and potential that aren't yet generally visible; or because they have a vested interested in the success of this model.
I've got a hard time reading these quotes. My brain automatically translates these to:
* Machines will replace humans, says The Machine Company CEO.
* Oil is here to stay, says The Oil Company CEO.
* Electric cars are a fad, says The Gas Company CEO.
They can't even be bothered to contract a thinktank to publish biased reports and give “expert interviews” at the news to support their views, no no. Just, “This is the future.”
I'd much rather see “This is the new company direction.” than nonsense interviews with high stakes gamblers that hope they bet on the right technology.
Maybe the important qualifier here is in “web developer” in the title.
That job title might simply become obsolete. When I entered the workforce around the turn of the millennium, there were people with titles like “webmaster” and “rich media designer”. Those don’t exist anymore. It’s not that the work has disappeared, but it’s been restructured and refined.
Developers and designers will always be needed, but the vaguely generic category of a web developer may be replaced.
In the future all jobs may go away, in the nearer future we may so become AI shepherds first training our replacement. ie where web dev jobs end prompt engineering jobs begin. At least that's where I'm transitioning, but not just prompting like using langchain and such for ai apps but I consider it prompt engineering or at least the dumber side of ai for those of us without masters in machine learning. I'm not sure if there's a name yet for what's essentially scripting and ai engineering without building the models etc...
> The developer job market is unlikely to ever fully bounce back.
> 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.
My prediction: As soon as they cut interest rates, probably around June or July of this year, firms will start investing again and the hiring will follow. Probably will have a sellers market for labor like during the pandemic.
What's going on now is more like a capital strike. Businesses aren't behaving very competitively, offering the same products and services as a few years ago just with higher prices. That won't last forever eventually this automated future will need to actually be implemented and many developers will be needed.
Most of the people in my network are gainfully employed and not having problems getting interviews. It's really strange. Feels like tech market is becoming like the dating app market.
> Management believes they can replace most of these employees with LLM-based automation
In the writing world, too. I’ve yet to hear of attempts at it being anything but a wasteful distraction. Probably works fine if you don’t care about accuracy or quality (and some use cases for writing don’t! Spam, astroturfing, paid trolling, mass-message scams) but otherwise LLMs aren’t close to replacing 80% of your content department. Maybe 5% at best.
But that doesn’t seem to have dampened enthusiasm for trying, yet.
One way to think about it: how good does it need to be for the managerial class types who are visibly salivating at mass layoffs to convert current professional jobs into 1099 “temporary” LLM quality control gigs? I’m betting we’re close to the level where some large places are going to try.
I do think it likely we’ll see a lot of organizations decide they just don’t care that much about (much of) their writing being any good.
Like rushing to replace your first-line support with a chat LLM: sure it’s terrible, but you don’t give a shit, so that’s fine (until you get stuck with the obligation to keep its promises…)
Things will get a little worse all over, GDP and productivity will go up, and we’ll claim it’s better. That’s not all of what’s been happening since about the 1970s, but it’s a lot of it, so, nothing new there.
How good does it need to be to get any passable result? Well, as good as the people writing it nowadays. The kind of quality you get from writers is not the same as you get from editors.
Or did you mean how good it must be before mangers actually start replacing people? That one I have no idea, and I don't expect it to have any relation with how good content people expect to read or how good content the LLMs actually produce.
All the CEOs he quoted saying programming is over are heavily or fully invested in AI succeeding. Was he not able to find other sources? Those don’t seem very reliable.
In the original dot comm bubble, there were about 700k US devs. Now there is nearly 5M. In the dot comm bubble, we lost 5-10% of developers. So far in the current environment, we've lost about 250k jobs, which is right around 5% also.
There was a big difference between now and then too: then, so many companies were nothing burgers with no real product and no real revenue.
I graduated in 2005. I chose to go into public education as a teacher because it paid more than software development (35k for a dev, 40k for a teacher).
Yeah, shit's hard right now. I still think it was worse back then. Nobody would consider it a better choice to be a teacher today.
I think this is going to turn out to be a great period for people to have started their own businesses instead of hunting for the same job (or worse) that they just get laid off from.
not if it's in ai, whoever gets to AGI first will cannibalize all other ai companies. Then AGI will eat every other software business. It'll be a chat interface that can be an ERP, accounting software. POS it can be and do anything you want it to do just by asking, so you only need one vendor. It's unlimited economic growth for once organization, death for everyone else. I'm still excited to see it happen, just a bit nervous what comes next.
I'm hoping for a star Trek society, personally. Though I just realized ai has always kinda been an end game... why doesn't it exist more in star Trek, sure there's data, but you'd think ai would permeate the star Trek universe. I'm a way to does in star wars, the droids have personality, self awareness, even self preservation.
I think that probably depends on just how productive AGI is. AGI is fundamentally resource-constrained just like everything else. It can only do as much work as there is power and compute infrastructure to support that work, which means that AGI will tend to be allocated toward the tasks for which it provides the most value. There will likely still be tons of things where the AGI is better than a human being, but because the AGI resources are better spent elsewhere humans still end up doing those jobs. As long as the AGI is resource constrained comparative advantage will still come into play.
This has always been my head cannon for why you don't see more AGI in Star Trek. The computers in Star Trek clearly perform a lot of work and automation how else can you operate such a large and complex ship with such a small crew, but they don't have unlimited resources so the AIs perform the work where they are most valuable and humans do everything else.
> This is despite the arrival of a boom in funding due to the generative model bubble.
Does the boom in funding for AI companies make up for the overall drop in funding from late 10s-early 20s heights? That would not be my guess offhand, which makes playing up the contrast between “plenty of funding” (for one specific thing) and “job cuts” feel a little overstated.
What does "web media" in this article refer to? Is it an umbrella term for everything non-independent, non-user produced media? Is a youtube channel "web media", is my blog, is huffpo, are the forbes not-independent-but-totally-not-endorsed-"sites"?
It was satisfying to come to the conclusion that unions can effect job security and then have that conclusion validated later in the article.
The problem is communicating the vitality of engineers in this symbiotic relationship, but the power dynamics and ownership models obscure dependencies.
“ Finding effective documentation, information, and training is likely to get harder, especially in specialised topics where LLMs are even less effective than normal.”
It's irresponsible to talk about the boom and bust cycles in hiring over the past ~4 years without mentioning covid and WFH [0]. Companies over-hired and over-retained for reasons like cheaper remote labor, prematurely extrapolating trends from large and sudden shifts in ecommerce/streaming/etc revenues, and ol-fashioned camaraderie (really). This positioned the industry to get doubly rekt, because even in good times, that retention debt would need to be paid down, and in bad times, it's even bloodier. There's some bullwhip effect[1] to this, where hiring too many people leads to needing to cut spend in a hurry, which invites more overcorrection, leads to needing to hire all the sudden, etc etc. It's my impression that these swings in developer hiring are continuing to go back and forth, with slightly lessening intensity each cycle, consistent with this effect. I think the pain will continue, but will also continue to lessen, over the next few years.
2. "has turned into a stock market driven movement..."
It's correct to tie a lot of this to market health and general macroeconomic environment. But what drives those things? The larger geopolitical landscape has a big influence here, albeit one that's hard to predict. For example the usa's post-war boom years happened when europe was in shambles and rebuilding, and the usa was mostly intact and well positioned for a manufacturing boom that serviced the rest of the west. there's some cause for optimism here, since again the US is recovering better that most other nations economically. Friends in JP and AUS are lamenting the weakness of their currency lately (or rejoicing, when they earn in USD and spend in local dollars). The next 10-20 yrs will see large, consequential geopolitical movement, particularly w/r/t china. Does the US successfully onshore/reshore semiconductor manufacturing? If so, does it exist alongside Taiwan, or do they weaken or face attack? What becomes of china and russia's demographic crises and budding codependency? Is the US drawn into a large conflict for those or other reasons? Does that cause the US to trend towards unity, or further politically atomize?
To me, these questions matter as much as, if not more than, AI advances. If the macro environment is strong, higher paying jobs will be more available, even if they're not needed as badly. If the macro environment is weak, AI successes will be a rocketship trapped under a low market ceiling. And if AI advancements are so staggeringly powerful that they mostly nullify the effects of macroeconomics and geopolitics.... this whole article goes out the window because the world it was written for has ceased to exist.
[0] Article mentions RTO, but only as a means of shadow layoff
It is impossible to take seriously an article that starts with "We have the worst job environment for tech in over two decades" and supplies no evidence.
He may be right or he may be wrong - it's great that we're having this conversation.
One thing that is overlooked here is that we're in the business of creating things to solve problems in the world. Code is simply the medium of creation.
Perhaps writing code does get scaled down over time and replaced with AI to a large extent. But the basic business and human needs that the code is solving doesn't necessarily disappear. Product management will still exist and there will be a need for humans that can talk to their human customers to understand their human needs. If the next step in that process of creation is prompt engineering and reviewing AI-written code, so what?
As creators and problem solvers, we need to always be reinventing ourselves anyways.
This will be a particularly painful and uncertain time for anyone whose job is going to look drastically different in the future, which is most of us. But I don't think that these arguments necessarily apply any more to software engineers than they do to a lot of other careers like writing, law, art, or movie production.
> If the next step in that process of creation is prompt engineering and reviewing AI-written code, so what?
1. It’s boring, soul-crushing paint by numbers scutwork
2. We value correctness and there’s no guarantee of it with LLMs
3. We value comprehensibility (answering the question “why did that happen?” in our systems) and there is an event horizon in these models past which we cannot see further.
4. If the jobs of software developers are materially under threat from this technology then no job that does not involve putting your hands on something or someone, or that does not have a legal/regulatory moat, is safe. Our society is in no way prepared for > 20% of its decent middle class jobs to disappear in 10 years time.
These are good points. Society will need to decide if we push back on AI and aim for upholding the standard of human quality or accept a higher efficiency at a lower quality.
I honestly think that a lot of the tech layoffs can be attributed to Elon Musk. When he bought twitter he started a social media campaign attacking developers as pampered, overpayed, etc. He then did the massive layoffs and twitter mostly kept running fine. Other CEOs took notice and have followed suit. Investors love it and have put pressure on the whole industry creating a trend.
It's basically an attack of capital against labor. Maybe we should have listened all those years when people kept saying we needed a tech union.
I don't think this is materially true, it's just a nice narrative. (And a way to entice investors, which, remember, is the CEO's job.)
> Most websites are buggy, inaccessible, and broken
It's been like this since at least 2010, so imo this isn't really anything new. We're just buggy in different ways. Switched from one footgun (jQuery) to another (React).
> That’s an entire industry removed from the web developer job market.
I'm not exactly sure what the "web media" sector is. Re-read that section like twice and I'm still confused. Programmatic advertising is dipping because things like FB/Google are losing out to TikTok and Instagram (& more tailor-made advertising). As someone that used to work in ad-tech, I don't really think those jobs are going away, just evolving in a post-Cambridge Analytica world. This has been in the works long before the LLM craze.
> Here we come to the reason why I did this research and wrote this analysis. I keep asking myself: what should I do?
The answer is always the same: start a company. I'm not sure why so many developers are terrified by the prospect.
> Sometimes all the news you have to tell is bad.
I dunno, I don't really see the doomerism here. I tend to be more optimistic than most, but even so, even if you can't get a job as a "web developer" (which really hasn't been a job title for about a decade now anyway), pivoting into data engineering, backend work, LLM work, infrastructure, etc. is not really that big of a leap.