AI has torched the market for junior programmers
seldo.com
seldo.com
The 22-25 red line had plateaued before the release of ChatGPT and was already trending downwards by the time ChatGPT appeared.
Additionally, it took a quite a while before vibe and agentic coding appeared and gained traction, and I cannot really see how the precipitous decline between say Jan 2023 and Jan 2024 can be attributed mainly to AI.
The "other" reasons mentioned later in the post seem much more convincing.
It will also be interesting to see 2019 through 2021. There was a glut of hiring post COVID and companies have to think about every dollar they spend post ZIRP.
And I typically work as a data scientist, but my role has involved more and more software engineering over time, so that might be driving some of the changes also.
This makes perfect sense and is a net good. There were a ton of awful bloodsucking SaaS startups destroying progress for these niches.
People who understand their niche best have taken it upon themselves to build exactly what they need.
The question about junior devs is a red herring. There are no "junior devs" because the title is obsolete. If you want to get hired as a dev, you need to at least show off some projects that pass scrutiny. This is the way hiring always was anywhere that wasn't a coding sweatshop.
We've always had people developing, in many forms. Scientists of all kinds, usually with Python, finance people with Excel, etc.
I think that yes, they can go a lot farther now. So this will make the bottom of the software curve grow 10-100x.
Now, the real question for developers is: what does this do to the middle and top or the curve? In my experience that's where maintenance comes in and anyone who's not a trained software developer (and even many SDEs) break their necks. "Casuals" will build what their need, but even with AI guiding them, it's still spaghetti.
It's going to be interesting keeping an eye on this, for sure.
This is just the natural next step towards a more mature kind of consulting. The client needs help scaling up their project or deploying it to the rest of their business. This is massive opportunity for any entrepreneur since the client is coming much better prepared.
I have several hobby projects. Is that really all it takes to get hired?
By "project" do you mean "product"?
What changed is tolerance of failing to meet that standard. That's how layoffs after overhiring happened. The bar was not raised, but moved back to where it was prior to the mid-2010s.
Anyway, it also depends on who is doing the interviews. If your "hobby" projects are built similar enough to what they have at that workplace, then the interview should feel like you're already working there for both you and them. Having absolutely no professional experience is going to automatically put you at a disadvantage. If you know of any startups, even if they're awful places to work, I'd much rather have a couple years of that on my resume than any internship.
What I was really saying is that someone who has been a teacher for a while and then learned to code with AI is now far more likely to get hired at an education software company than someone fresh out of college with a degree in CS. That's the way it should be! That's genuine progress. They may not end up in a coding role, but they'd absolutely crush it as a project manager.
Huh. That's my situation... Sort of. Been programming since I was a kid but went into another field. That gave me hope, thank you.
And as systems become more complex with time, we will need more people with the title “software engineer”
I as a senior am still called in to consult and shoot down attempts to actually integrate it into the wider system without triggering a full technical review.
That said, I find a lot of anecdotal information from many people in the space that tech was flooded by a lot of junior programmers who were basically in it for the money with minimal training and they're having a hard time of it. The same thing happened to webdev during dotcom.
Also Fable 5 isn't "that impressive" as a lot of people have that kind of intelligence since 6 months+ by using combo of models and loops (I scored better on HLE than gpt-5.5 xhigh last January with some good tooling and 6x the cost), but for a lambda Claude Code user, I can see why it looks that good.
On the bright side, maybe that means the end of new javascript frameworks every 6 months :)
However, the ability to reason generally on novel problems is AGI and we aren't there yet. Eventually in the absence of AGI, we will have to train models on them, and that will require data.
In my experience with all agents, including Fable, is that they work great when there is automated validation. But as soon as it needs to design something, it just keeps adding so much slop.
End result - 2 hours later it produced a convincing theory with lots of references, and burned a bunch of tokens too of course. just for fun we tried its suggestions and deployed them to prod. Guess what? Didn’t fix the issue. Alas, a human was needed after all.
either everyone’s working on toy problems, or they’re working on very cookie-cutter code. I’m really not sure. I DO remain impressed with Fable 5 but the idea that we’ll all be unemployed in 2 years is hilarious delusion. we’re already at the point where many organizations are scaling back some of their AI spend.
And while we’re talking about hilarious delusions, perhaps you should look at the current capability curve of AI and weigh it against the constant stream of arguments for why it couldn’t have continued at every point and yet has.
Yes, the solution is to just burn more tokens.
In this case i had no empirics in the loop. The scenario was only reproducible under high api load. I could load test, but management isn’t eager to spend prod-like costs in staging (requires scaling opensearch a lot in stage). What can i say.
"Models can code well now but they cant do high level architecture" is just a logical fallacy. Its literally only true in this particular moment in time. But if they can code well, whose to say they wont architect well? And at that point, what do SWEs do? If anything, SWEs are in the critical path of automation for these AI labs anyway, so theres a very strong incentive to automate us out vs other professions, and it'll happen soon. All these random 1-off datapoints of "Fable 5 can't do X very idiosyncratic thing" are completely missing the point. 6 months ago, even attempting that problem with any "tool" would be totally intractable, and now it _just_ writes a slightly subpar solution. You can do some basic extrapolation here, its not that complicated.
Your best bet is to just chose a different career, or, if you still want to be in the software industry, be more enterprising.
I’m at 90%+ code AI generated by stats. I work in embedded systems. It still goes off the rails all of the time and needs a heavy hand to guide it. It does not currently feel like it will ever be truly able to operate independently. It’s a very useful tool, but it’s just not there yet in my day-to-day.
Obviously, YMMV.
Maybe your point is just what the difference is to me. It’s competence. On hard problems Fable can just iterate by itself like nothing before. You give it a task, it can plan it, formulate and falsify hypotheses, and do more complex reasoning than anything before it. The main difference is simply that it gets more right and can go deeper on everything. Yes, you do still need to course correct it, but it’s just on the next level.
It’s like what everyone experienced when Opus 4.5 came out last year, and the penny finally dropped that AI coding had arrived. If these sorts of jumps happen even one or two more times then it’s all over. It might even be over now and it’s just too expensive to run, but that will change over time.
If you can, you should try it and see for yourself.
Honestly, the biggest improvement AI have brought me is the complete end of my imposter syndrome. People say comparison is the thief of joy,but in my case, previously invisible devs getting more visibility, suddenly pushing new code every week at the same pace as me but clearly without any understanding any of the underlying issues did wonder to my self esteem. I still consider myself among the worst of my group, but my group evolved from 'all devs I know' to 'the competents', which is very nice.
To the direct point: I've used AI for a lot of things. It's good, you can definitely build some things faster. It's far from perfect.
Also, if we were sooooo close to AGI, why would Anthropic and OpenAI continue hiring SDEs like mad, which they are? Why would Claude & co have embarassing bugs years after release (the glitchy scrolling bug, their websites being fully synchronous in 2026!!)? Etc.
The steam engine was invented in 1765 (+/-) but employment crashed in the textile industry starting in 1815, 50 years later.
LLMs were practically invented in 2017, 9 years ago.
Of course, software is software and we live in a faster age, but guess what, human needs and desires are infinite so requirements keep going up.
At this point I have 0 trust in purely agentic coding (i.e. the 0 supervision approach). YouTube randomly stops displaying subtitles, Gemini CLI (before its shutdown) and Claude Code suffer from constant refresh glitches, the Gemini and Claude mobile websites are synchronous, like we're living in the 60s (if I minimize Firefox both Claude and Gemini stop answering and Gemini until about 1 month ago would also lose my prompt - this is the kind of thing people were fired for in the past, etc). If this level of buginess is the best agentic coding can produce with unlimited state of the art mega corp tokens, software itself will collapse faster than the job market.
Claiming that there will be no more SDEs of any kind, worldwide, in max 2 years, is an extreme position.
I'm not even an SDE. If they're turkeys, AI boosters are pheasants :-p
And I know why the legitimate AI boosters do it. They were genuinely impressed by what they saw in agents, especially since many hated programming anyway. They extended the line into the future and assumed complexity was linear. These people looked at LLMs and agents and didn't see what was there, but what could be (frequently from a position of major SciFi consumers through their entire lives, as all techies are).
Optimism is a positive, extreme optimism is a great flaw.
Nobody said that. The claim is that within 24 months the models will have the capability required to replace all of us. I stand by that. How quickly that moves to everyone actually getting replaced is a social and economic question. There may very well be well be a long tail, I’m sure today you can find traditional weavers and barrel makers somewhere, they just happen to be economic novelties.
For the record, I’ll say that by 10 years out the profession is hollowed out enough for it to count as destroyed, and the process has already begun.
LLMs/turtles all the way down?
Turns out that as admitted, the opposite was true and was predictable. Such that, in late 2024 [0], I predicted that there would be more layoffs in 2025.
> In March I checked in and found startups substituting compute for labor at record rates, with the wave of new jobs nowhere in sight.
Of course they would. Why hire a junior software engineer when you can replace them with an offshore remote mid engineer at 1/10th of the cost and give them Claude?
Surely that makes all of this even cheaper? False. Just ask Apple. [1] Or Boeing [2] [3] with their expensive offshoring and their trade secrets either leaked or the quality degraded.
And those drunk on token usage are now limiting it because it is expensive. Ask Meta, Tesla, Amazon and Microsoft why they are not "tokenmaxxing".
[0] https://news.ycombinator.com/item?id=42490692
[1] https://www.reuters.com/world/india/india-investigating-tata...
[2] https://www.lemonde.fr/en/economy/article/2025/06/12/boeing-...
[3] https://www.computerworld.com/article/2513787/boeing-and-the...
Anecdotally, I know one couple who is selling their house to move back to a tech hub for better job opportunities. Another couple working from a rural location recently had one partner lose their big tech job. If the laid-off partner doesn’t get a new job in the next few months, I expect that they will start talking about moving, too.
That's not the explanation according to this article [1].
> Or they lost their job while living elsewhere and had to move to the Bay to find a new one.
It is hard to afford Bay Area rents -- even if you have a job. I can't see someone looking for a job signing a 1-year lease at these rates. Also, unlike the old days, interviews are remote, being physically present in the Bay Area doesn't give you an advantage.
[1] https://www.sfchronicle.com/bayarea/article/remote-work-stab...
The article you linked to seems to say only that employers have been unsuccessful at enforcing 100% in-office work. It does not say that people have been successful at keeping their 100% remote jobs in other states and far-off locations.
Overall, we seem to be reaching a compromise where many are forced to live within commuting distance of their office but aren’t necessarily there every day. This compromise is what is pushing up rents from what I am seeing among my network.
In other words, “remote” and “hybrid” are not at all the same but you are using the terms interchangeably. You can’t live in Idaho and commute to SF three days a week without major inconvenience and expense.
Maybe you aren’t aware, but house prices skyrocketed in many locations (like Idaho and Florida) during the pandemic because of an exodus of Californians who were told that they could keep their high-paying jobs and live almost anywhere. The tide has reversed now.
As for interviews, I have read multiple Hacker News comments from people saying that their solution to rampant AI cheating has been to make interviews in-person again. So I’m not convinced that you’re right about that either…
80% of them should reconsider their career paths. They glaringly cheat during interviews, they can't answer basic software questions, and they're clearly in it just for the money.
They were never interested in software. They just saw the success of software engineers that put in a lot of work, and were fooled by the 2021-2024 hiring spree to think it would be easy.
That's often how jobs work. I happen to like my job, but let's not pretend like everyone is going to make sufficient money following their most passionate interests.
I do the job for money. Period.
I also happen to really enjoy computers, systems, etc. I enjoyed most of the jobs I've had, but the reason why I was there doing the job, as opposed to being at home nerding out for my benefit alone, was always the need to pay bills.
So it really is good to find people who are a good fit for a specific position - and who, in the worst-case scenario, would do the work even for free or without anyone supervising them.
The question is whether the younger generation has grown up with the motto “money first and at any cost”… Not everyone, of course, but when I see how everything everywhere is reduced just to monetary value, it makes me feel a little down sometimes.
Housing in particular is a driver of people optimizing for cash.
For sure, sufficient money is important part of this equation.
But I strongly believe that if you find (your sort of) fun in what you do then you become faster and more efficient at it.
Edit: Cleaning, nursing etc. Yes there are professions that almost no one wants to do. But that doesn’t mean that these people do it just for money.. this is prejudice that if you don’t like it and don’t have fun that other people are the same. No, they aren’t.
Some, actually many, jobs are just shitty.
Anecdotal in my experience, but that also used to be the case for most CS majors back when I graduated... but that was well before the rise of the FAANGs.
Or they will refuse to self invest, which makes them a worse long term investment for a company.
With that sort of carelessness in the quality of your sentence, I probably would not hire you.
Seniors and above lead by example.
Great, when is he applying?
I can't imagine what the ratio is now after everyone was pushing their kids into coding and every douchbag chasing a high salary tried to enter the field. Maybe close to 90%?
I can't imagine there is much of an issue for bright, passionate folks just starting their career if they can manage to communicate their passion and skills to a company. Sure, they might not go directly to a fang, but there are literally tens of thousands of companies who staff engineers and if I were new that's where I'd target to get started.
The reason you bother with them is and has always been to create future senior engineers that understand your business well enough to ensure business continuity.
Too many if not most can't even change a single sentence in their curriculum in 1-2 years.
It doesn't mean University isn't worth it, it's worth less if you aren't self-directed learner building things.
Maybe, just maybe, companies should invest into training again instead of outsourcing training to universities and saddle the prospects with the cost in the form of student debt? FFS we used to tell our children "if a job requires you to pay for entry, you're getting scammed" - and yet, we've all accepted it with "academia", "coding bootcamps" and god knows what else.
Universities should be a place for the gifted to advance science, not be degree mills for large companies too goddamn lazy and penny-pinching!
Extensive training made a lot more sense where people stuck around for 15-30 years but I expect a lot of people here would find that ludicrous.
There's nothing preventing a company from saying "hey, we're paying for your training and expecting 5 years of service in return, and in exchange for this, you get guaranteed 2-5% above inflation".
And if government would have some sort of sense, they would force companies to act like that again.
But nobody wants to set up programming trade schools or apprenticeships so shrug.
They weren’t a panacea either.
Universities aren’t vocational schools. An undergraduate education can, at best, teach you how to learn complex topics independently and give you foundational knowledge you can build upon later whether you’re going into industry, pursuing a PhD, or doing something unrelated.
The place for you to learn “practical” software development is an internship or an entry-level role, assisted with a lot of self-directed learning that hopefully university made possible for you.
They also universally advertise themselves for many programs that are current.
The "teaching you how to think" part isn't out of line at all, however, applying it in the real world has to be taken relative to what students are explicitly told before registering in university as to the outcomes they will achieve.
I was a TA in grad school and did some tutoring and some students just didn't have the math basics they needed. I could help with a few things but couldn't really teach remedial high school math.
One might argue that it might not be a bad idea to have apprenticeships like they do in traditional trade crafts. I'd argue that this is what the junior programmer role is.
Junior programmer roles are not internships or apprenticeships at all. Internships before the students graduate are internships.
Instead, asking an employer to act as the employee's parents, and pay to educate the new graduate so they can disappear in 1-2 years makes an increasingly harder choice.
This is part of the reason that Junior dev positions were not succeeding, prior to AI and the pandemic.
in your place i’d focus on delivering big wins for end-users via multiple roles ; i personally believe the days of being a specialist in one type of stack (ex: “i’m a ui developer”) are over , in this next stage we’ll say “i build products for young people looking for educational opportunities “ ; ie you’ll beed to cut across use-cases and disciplines in a way that was atypical just a few years ago
It's a job, working for someone else. What other reason is there?
Remove the need to pay bills and you'll be left with no one applying for a crappy job making money for someone else.
Implementing the craft has nothing to do with putting up with employment and employers, workplace harassment, waking up at unhealthy hours, wasting most of one's time awake doing things for someone else...
our noob to hero pipeline these days is just requisitions that cover some basics but flat out say its an entry position and a chance to step into something new. we quickly weed out the overqualified and find candidates who seem like theyre genuinely just looking to find a way to break into something new, and our interview process is largely centered around getting to know them as a person what they're all about and we seem to do an okay job triangulating "this kid is curious, seems like they'd glue well with everyone here". mildly grill on some technical stuff but mostly just to get a read on where they're at & make assessments on what we're willing to teach. we actually don't do whiteboarding coding exercises or any of those challenges, not really a fan of those and i've never felt like they were a useful litmus test on whether or not someone will be successful in a role. people can hide a lot of insufficiencies/ego/insecurity/toxicity behind coding exercises, polished resumes and mastering interviews. it is of our opinion that the journey to expertise/excellence also involves having a good grasp on who you are as a person & who you want to be, and a solid moral compass you can navigate with. because there will be highs and there will be some very low lows, any journey to excellence sails these seas & we're very interested in people who can be open and honest about where they might be.
at this point we've reasoned out that the normal requisition is just going to get an influx of people who think they're charmers and can rehearse an interview and then pan out to be a nothingburger. we've had a couple of those in the past few years and it's annoying. we don't hire often do but when we do we're just interested in having someone around that we like and seems eager to learn, we've had great luck with this formula and this seems to pluck out people who get lost in a sea of incredible looking resumes, we give them a good learning track and goals and they seem to leapfrog past every goalpost we've put in front of them. as far as juniors go, we like hearing about other weird non-computer problems they've solved in life. when we find the right candidate we kind of just know when we talk with them, they universally are pretty open about their own shortcomings but just demonstrate some sort of very passionate need to build or solve things and find a community where their contributions are valued. we see the mission as needing to build them up as a person first and the nerdy stuff is the fun sidequest they can join us and chase dopamine with. we all enjoy teaching & watching people grow so it works out pretty well and we've transitioned a couple to senior positions in the past 4-5 years. people who, when we first interviewed them years ago, may have not had any business on paper being in this field. people we're proud to have watched pan out to be incredible resources, some of the heaviest hitters our org has seen, and if im being real at this point really great friends.
How so? I am curious what cheating you are experiencing and how you detect it.
I'm curious what they do to cheat during interview?
Imagine my horror when I found out recently both of them have a computer science bachelor's degree from a UC Santa Cruz.
Just look at the writing on the wall, there will be no need for senior SWEs of any type within 1-2 years anyway, and shortly after that we won't need Staff SWEs, etc. People here are way too myopic. AI is progressing very fast. We went from hiring juniors in droves 3-4 years ago to basically proclaiming the death of junior SWEs. Who is to say this won't continue up the ladder?
AI will be good enough to replace all SWEs in any capacity - there is no point in "investing" in rebuilding this ladder when you can just invest in more GPUs (in the case of oai/ant/meta/google/etc). or just pay those aforementioned companies more in tokens if you are a smaller outfit. The cost effectiveness of those tokens will only get better over time, until they are competitive in cost : intelligence when compared to any human SWE.
> The jobs disappearing are the ones where the work product is code written to spec. The jobs growing are the ones where the work product is judgment about what code should exist.
AI is happy to follow instructions, no matter how stupid, unoptimal or unnecessary those are. To be successful, you need someone to understand the details and make the decisions.
And while that "someone" could be a person that does not go in the details and doesn't understand the code, they would be equivalent of non-technical CEO - sure, those exist, but they have a much harder time creating successful products.
And how many people do you need for this? There are many roles where people are literally hired for their programming/engineering skill. Modern LLMs _largely_ commoditize that skillset.
There are not that many novel things to do, and even if there are, you don't need too many people to do them as LLMs give you more and more leverage over execution details.
I am currently work on embedded / low-level / devtools and there are tons of novel things to do. Perhaps the "programming" skills are not needed as much, but demand for "engineering" is as high as ever.
Maybe it's the case in web dev? TFA shows that the number of "web developers" is decreasing strongly.
Typically for a project you’d have something like 1 senior, 1-2 mid and 2-3 juniors and sell the team to the client.
The junior/mid is where the margins are, as seniors knew their value and commanded a bigger salary with little margin for profit, but juniors aren’t paid as much, yet you can still comfortably bill the client.
Nowadays it’s 1-2 seniors for the whole thing and the service company is expected to pay for the tokens the seniors use to replace the lucrative juniors, so it’s a double hit for the company.
The people who are succeeding are learning and playing and building their own experience they can demonstrate.
There's few shortcuts if any that last. It comes out in the wash quicker.
On the senior end there remains a gap and advantage between understanding of human vs understanding of AI on how best to approach or work through things.
Used to have an “academy” where we pick people up from first year in uni and get them to learn from zero.
Haven’t had one in ~3 years, no plan to.
Also half the company was downsized.
No hiring in 3 years, cutting tons of people and tons of people leaving, we’re now at a point where in the tech i work in, our most junior person is a senior with about 10 YoE
All our good ideas come from senior level people drawing from a rich history of experience and bold vision.
While experience is important, it also creates blind spots as it's quite hard to have every experience about a particular topic unless you have been working on that tool/stack alone for decades.
Doubtful.
You'll be expected to produce much more, you aren't likely to see raises commensurate with the increased productivity, and the AI tide is still rising and will likely eventually put downward pressure on your pay at an increasing rate.
Casually pretending that decades of software engineers comparing programming to typing and saying "you are not a programmer, that is not a job" long before chatbots, didn't exist.
The job was always "software architect", "software engineer", "web developer", etc.
Juniors now need to be more skilled than before
Why? they didn't know how to copy before or something?
> spin up and manage dozens of juniors
Let's not call it a "junior" if one trigger word deep in the filesystem can make it forget it's a programmer and start building Minecraft castles. Your comparison is unfair to anyone with even 1 week of programming experience.
LLM are so powerful exactly because they can interpret unstructured inputs in so many ways, prompt injection is a feature not a bug
The same way "computer" and "calculator" stopped being a job title when they became devices might be a better way to reason about what is happening.
A junior developer will still be much better at vibe coding (in the technical sense) than a senior manager who's never looked at code, but perhaps not as good at choosing what to code in the first place.
Agreed. At one point, for this reason, they introduced graphic calculators into the classroom, for example, circa 1990, the TI-81. It became more important to understand how they worked than spend the time to do the calculations.
A couple days ago, I wanted to understand how the Contrastive Language-Audio Pretraining model works so I spent 3 hours making data visualizations using Claude Code. [0] Look at the math notation which to me is very intimidating. Nevertheless, being able to visualize the values as animated cells in a grid synced with computation, I immediately grokked how the information was transformed into something that can be used to train a model from audio data -> image data -> 512 embedding.
Working through these problems in 3 hours using a coding agent will be analogous to working through calculous and linear algebra problems in 1990 with a graphing calculator. It is still 3 hours of wracking a brain.
1, "software engineer" or even "programmer" is not becoming a device is it?
2, those are devices that operate for specific purposes. This is a general purpose autocomplete, and therefore if you support his logic you should also say now with LLMs "writer stopped being a job title", "assistant stopped being a job title", "teacher stopped being a job title", "manager stopped being a job title", ......
3, result of their calculation is deterministic and only depends on their wiring, not original work of humans. Imagine if calculators started to degrade when humans completely stopped doing maths by hand;)
If it can be taught, an AI can do it. The only work left is either manual or inherently new.
Absolutely not the case now. Maybe it will be in future, but that's basically impossible to predict.
So you could have the senior oversee the junior or just have them oversee Claude.