I just read a stat that made me sit up straight.
By 2026 (well, it's happening), ninety percent of all code is predicted to be AI-generated.
Not ...
For further actions, you may consider blocking this person and/or reporting abuse
AI isn’t replacing all developers yet, but it’s absolutely changing what “developer” means. The value now lies less in typing syntax and more in understanding systems, context, and making good decisions.
I feel it's a bit like when you want to build a house extension. Maybe once upon a time you'd have done it all yourself, but now you hire a brickie, plumber, plasterer, electrician etc. Your job becomes that of architect. But if you've never done all these other jobs, how can you tell if youre getting what you wanted?
AI is a bunch of skilled tradesmen, but right now it sucks at being an architect. Only a skilled programmer can tell if the result is fit for purpose. And I can't see that changing any time soon.
One of the most hard-hitting articles I've come across in quite a while - good article, which asks some tough and deep questions, but quite gloomy - I'm pointing out this sentence in particular:
"Developers are now applying for 200-300 jobs just to get one callback"
That's probably the most scary sentence in the article ...
Compare that to a decade ago, when companies were scrambling to find devs - anyone with some decent 'coding' ability would be snatched from the middle of the street, and put behind a desk as a developer for a comfortable salary - the job market for devs was a "seller's market" - I think it was the "golden age" ...
Would all of the decline be attributable to AI ? Probably a large part of it, but I have the feeling somehow that it already started before AI came on the scene - but that AI accelerated it, a lot ...
Two years ago I wouldn't have thought things would evolve (some would day "deteriorate") THIS fast - it's far exceeded my expectations ...
So yeah, these are big questions, I would almost say philosophical ones - what do "we" want to be?
If right now you'd be 16 or 17, leaving high school, choosing a discipline/subject for college or uni - would you still choose "development"? Maybe the 'old fashioned' hands-on blue collar jobs (plumber, bricklayer, you name it) are going to experience a big 'renaissance' as they're much harder to replace by "AI" ...
P.S. an antidote to the gloominess and pessimism - I came across this article, which has a far more optimistic take on the subject:
dev.to/j4s0nc/the-bloated-kitchen-...
And this is what I sort of see as the answer - also to the question of the junior/mid level dilemma: I think what "boot camps" etc should start teaching is indeed "systems thinking" and "big picture thinking" - forget about coding bubble sorts etc etc, use 2 or 3 days to teach the basics of coding, then switch to teaching "the big picture":
Backend vs frontend, cloud/devops, what is HTTP, what is a database, why is security important - but only the basics - don't even bother teaching the finesses of SQL or whatever, all of that can be picked up "as you go" ... chefs, architects, system thinkers - even "juniors" need to be just that.
Hey leob,
Thank you very much for reading my article and for the kind words about the "antidote"—I’m glad it offered a different perspective.
The era you called the "golden age" was also the time when the barrier to entry collapsed. We ended up with an oversupply of developers doing low-value, high-churn work, driven by "tech for tech's sake" instead of business problem-solving. This created massive, complex, and ugly codebases—the "bloated kitchen."
AI may be the necessary shock that is ending the era of mediocrity. It can handle all that rote work instantly, forcing the market to focus on what actually matters.
I agree 100% with your solution for the next generation:
Bootcamps should teach systems thinking and big picture thinking. Backend, DevOps, security fundamentals, and what a business problem is.
The future isn't about writing code; it's about architectural intent, judgment, and curating the output. These are the higher-level skills that should be focused on, and they are the skills that will hold value.
The job market maybe painful right now, but is it a necessary correction that will bring the focus back to craftsmanship and actual value? I hope so...
Thanks again for connecting the dots between our posts and for fueling this conversation.
/j4s0nc
I've loved the act of programming since I started doing it in the mid-80s. But even more, I love getting stuff out there that helps others do their jobs better, get past their pain points, etc. So to me, shipping is king. And AI helps me ship faster, which is a win in my book.
You can still be programming in 10 years -- but you might be doing it just as a hobby.
We are the blacksmiths of our era.
have you tried those tools ?
I have
This is a bold assumption, but it's possible that we are generating more code, and that 90% actually refers to a percentage of the newly generated code that doesn't necessarily fully overlap with hand-crafted code?
On a side note, I haven't seen anyone, historically, tracking metrics on how much code was written (when it became available and popular) by autocomplete, auto-generated/refactor tools by IDEs, or copy-pasted from SO, as opposed to being manually typed word-by-word into a simple notepad. AI is just a tool, it does some of the work, not all of it, and it doesn't operate autonomously. Ultimately, it still requires an operator, just like an IDE.
Regarding the LinkedIn part: Firms were using algorithms for screening CVs long before AI, among many other less-than-ideal practices that perhaps should concern us more than AI usage. To keep this concise, I believe the recruitment process is broken in this day and age. It was broken before, but with today's scale, edge cases are no longer so edgy.
As for the junior replacement part: Again, this is just my opinion, but every time people compare AI with junior developers, I feel it's truly disrespectful to junior developers. I've had and continue to have juniors on my teams, and they bring a value that AI can never replace. They offer fresh ideas from a constantly changing world perspective, possess the energy to be bold and try new things, and accept failure better than some seniors, who might feel their ego hurt just by admitting "I don't know." They have potential, in contrast to AI, which is merely a mediocre (in the most literal sense of the term) aggregator. Any company that currently believes it shouldn't invest in junior talent will likely be plagued by high turnover and monstrous projects that nobody understands how they work or why they break, again, mediocrity here, which maybe is where the so called "professional" software is aiming for idk.
Ultimately, AI is just a tool, neither good nor bad. I have seen really great code, but also some of the most bad, spaghetti, and wtf code in my career, before any AI was doing anything, the only thing that change is speed imho. Our job is now rapidly evolving again, as it has many times before.
Where did you get the 90% figure from? It seems very optimistic even for an AI fan.
How is it doing that? It is not thinking about new solutions, it is generating old solutions.
I wouldn't trust learning to AI. It is more like a librarian that knows where to find the information you are looking for. And sometimes it just fabricates information out of thin air.
I think the biggest problem is that people are spending a lot of money on AI now. But we already seen the cracks in the good news show. What will happen when the AI promises are starting to fail or beginning to create more problems than it is worth. Who is going to come up with the billions to pay off their debts? It is not like the banks where we need to store our money.
More productive + massive technical debt describes the same mistake many managers have made for decades: ship shit faster, move fast and break things, minimal viable product, or whatever they liked to call it. Now AI is the excuse for caring even less for code quality, security, usability, and maintainability. That's not innovation. That's the opposite of innovation. Not AI, but people's decisions are destroying traditional learning paths, if there will be less seniors in the future, it's our fault as an industry. After the burst of the current bubble, current AI will just be another generation of helpful tools and resources, like IDEs, linters, and StackOverflow used to make coding faster and more comfortable in the past decade. Companies will need more human developers again eventually.
My take: personally, I'm already a senior, working with AI and far from satisfied with its lacking code quality. As a senior, I'd probably learn to code as a hobby, contribute to open source or create other kinds of non profit projects, while working other jobs in the meantime. That's exactly what I used to do for several years when I was young by the way. Whatever is your take, good luck and don't believe the hype.
AI isn’t killing devs; it’s killing shallow dev skills. If your value was boilerplate, you’re replaceable. If you can reason, design, and judge code quality, you’re fine. The real crisis is the broken junior pipeline, not mass senior unemployment. Adapt fast or fall behind.
I use AI on daily basis and it is ultimately helping me being better developer. But I still think many "predictions" are exaggerated and will turn inaccurate. People like to hear doomsaying. It sells better than sober-minded analysis.
I have recently read an interesting question: If AI generates so much new code so quickly and it is so helping devs shipping faster, where is the spike in the new software being delivered? Did you noticed any when you think about it now?
The effect of AI seems to shift us from writing the code to understanding it and moreover - understanding what are we doing and why. And this is a good thing. I am somewhat afraid of the future, because the economy struggles and war is knocking on our eastern door (I live in Czechia). But I am not afraid about the future of programming and my carreer.
Just yesterday I read an article about the current state of a thing called "vector databases". Have you heard about them? They were predicted to kill traditional database systems because of being so much more clever. Pinecone company got 750 million dollars valuation in 2023... The reality check two years ago - it is not that easy. It was an evolution forward, but not such a radical revolution as once predicted.
And this happens to most of the IT industry buzz-words I had chance to witness in the past decade I am part of the show. "Agile", "cloud", "DevOps", you name it... All of those brought significant leap forward, but also accompanied with a number of steps sideways, none was an ultimate solution and all introduced new problems and challenges to tackle.
I too am conflicted. I think I had entered school for tech/ graphic design at the worst time a few years ago and have been burnt out, but eager to be creative again. I didn’t like my experience with chat gpt trying to connect it to GitHub, although I have found smaller models like Roo and it’s almost hopeful as a solid tool as there is like code teacher mode. Great article, thanks for the honesty there.
90% is an eye-catching headline number, but (even assuming it's accurate), what does it actually mean in concrete terms? "AI-generated" is very vague. For example:
My exact keystrokes (other than selecting from suggested options) were as follows:
I a wr t sent on m p usi au compl, wh pe wo y s is "AI-"?Interesting take. I think the junior dev concern is real, but maybe not for the reasons people usually point to.
The issue isn't that AI replaces entry-level coding. It's that the whole career ladder used to depend on juniors starting with simple tasks, learning by doing, and slowly leveling up. If AI handles the simple stuff, how do juniors build foundational skills? You can't start by "directing AI" if you don’t even know what good code looks like yet.
The mid-level compression you mentioned is scary too. If companies think they can just have senior devs with AI instead of building teams, that pretty much wipes out the path from junior → mid → senior.
Maybe junior roles need to shift to reviewing AI output, debugging weird edge cases, understanding trade-offs, instead of cranking out CRUD endpoints. The skills that matter are shifting from "can you write code?" to "can you judge code, spot issues, and make architectural decisions?"
And yeah, documentation of your actual contributions matters even more, because "I prompted AI" isn't a career story.
How much financial data processing, calculation, and reporting over the past decade has relied on Excel, Google Sheets, or Numbers? I'd confidently estimate it's north of 99%. A person relying solely on pen and paper for complex bank financials today is truly one in a million.
Yet, did Excel replace accountants? While it certainly made some manual roles obsolete initially, it ultimately became a minimum requirement for the profession. Accountants didn't vanish; they evolved to handle higher-level tasks and analysis that a spreadsheet tool cannot do.
The same trajectory awaits coders with AI, LLMs, and agents. These tools will elevate the floor, not eliminate the ceiling.
It reminds me, before I started using AI, I was good at coding, especially in C and Python. But after AI became widely used in my country, I got used to relying on it. Now, when I try problems on HackerRank or LeetCode, I feel like my skills have gotten worse. My question is: should I learn from scratch again, or just continue using AI? This question has been on my mind :)
You can continue to lean from scratch for yourself but if you have tasks and you are obligated to use ai use it
Yeah, that’s what I’m doing rn, thanks for reminder bro
You're welcome
My two cents : AI will drive costs down for a while because it's faster.
But some devs will continue to code manually, learning and writing quality code. These will become "luxury" devs, like luxury brands, they will not make money with volume but with quality. And not everyone will be able to afford them.
So, in the end, two ways for software dev : the Temu way, cheap (both money-wise and quality-wise) and fast, or the Vuitton way, expensive and slower to make, but high quality.
What about the situation where funding stops fueling AI adoption and real AI costs hits the market, leveling things at midpoint?
People needs to adopt to the new way of working and ensure to upskill their AI skills to optimize their work and be more productive.
Thank you very much for the interesting article.