I do think most of us can attest to the fact that, entry level roles are not 'entry' anymore, I do see alot of tech jobs on sites with '2 or more years of experience' (how are we supposed to get that experience if no one wants to hire entry level applicants?!)
I check my email like checking the time, just to see if I get a 'congratulations..' message, or an opportunity that would help pivot my career somehow? It was a saturday, I didn't sleep much the day before so I woke up around mid afternoon. First thing I usually do is to check my mail (even before other social media apps). I saw a 'congratulations, you have been shortlisted for the role...'. I was so excited! Till I went for the interview and did not get the job...
I Applied for a Part-Time Junior Role. I Didn't Get It, And I Realized My Degree is a History Lesson.
Between lectures on compiler theory and trying to master DevOps, I just wanted a part-time junior gig, something to pay the rent and finally get some "real-world" experience on my CV.
I found a startup nearby hiring for a "Junior Web Developer." The job description was standard 2024 stuff: React, Node, basic Git. I’ve built a dozen projects like that. I walked in thinking I was overqualified.
I walked out realizing the job I was looking for doesn't exist anymore.
The Interview that Broke Me
The interviewer didn't ask to see my GitHub. He didn't ask me to whiteboard an algorithm. He just pointed at a screen with 2,000 lines of fresh, AI-generated TypeScript.
"I just let go of our last junior," he said, and my stomach dropped. "He was great at writing code, but I don't need a writer anymore. I have an agent for that. I need a Forensic Auditor."
He set a timer for 20 minutes. "The agent says this payment gateway refactor is 'Successful.' My logs say otherwise. Tell me why the machine is lying to me."
I sat there staring at the most "perfect" code I’d ever seen. No typos. Perfect indentation. But I froze. I spent my labs at Uni learning how to create loops, not how to find a microscopic logic flaw in a "perfect" hallucination.
I didn't find the bug. The timer hit zero. I didn't get the job.
Why the "Junior" Label is a Lie in 2026
Walking back to the Mile End station, it hit me: The "Junior Developer" hasn't just moved, it’s been deleted.
In 2022, your value was being a "coder." In 2026, your value is being a Senior-level Filter.
The industry has lost its patience for the "ramp-up" period. Companies aren't hiring for potential anymore; they’re hiring for control. They want people who can act like a Lead on Day 1 because the AI is already handling the "Junior" work (boilerplate, unit tests, basic CRUD) for free.
The Computer Lab vs. The Real World
At Uni, we’re still arguing over semicolons. In the real world, the Lead Architect is managing a fleet of 10 AI agents while they sleep.
The "traditional" student roadmap is a total trap. If your resume says you "know Python and Java," you're competing with a calculator. If you’re still grinding LeetCode Easy, you’re training for a race that ended two years ago.
To actually get hired in 2026, you have to leapfrog the "Junior" phase entirely. Entry-level now means:
System Forensics: Can you debug a "perfect" system you didn't write?
Orchestration: Can you manage the agents instead of being replaced by them?
Architectural Judgment: Can you tell the CEO why the AI's "efficient" code is actually a security nightmare?
The Dark Reality
We are the first generation of developers who have to be Seniors before we’re allowed to be Juniors. It’s a "Programming Dark Age" where the bridge from student to expert is being burned down by automation.
I didn't get the role because I was a student trying to be a coder. I should have been a student trying to be a Lead.
To my fellow students: Are you still building "Weather Apps," or are you learning how to audit an AI's hallucination? Because the industry isn't waiting for us to catch up.
I know this hurts, but we need to talk about it.
Top comments (123)
What a shit exercise. I don't think anyone can find a bug in 2000 lines in 20 minutes.
The way I would approach it is to ask for the previous code, to check what the refactoring did.
The command that triggered the refactoring, to check if there is language that AI can interpret wrongly.
Look at the logs, I assume they are the actions the AI executed, to check if a wrong command is given to AI by the agent.
If there is a bug, there is always a cause. When people produced the bugs it was easier, because they are typing slower. And if you know them, you understand were they fail.
With AI in the mix there are different stages where it can go wrong, so it is much harder to debug.
While syntax is a important part, schools that provide an IT trajectory should also teach the theory. Only learning syntax is something for a bootcamp.
There are smart people that can understand concepts very fast, but most of us need time to learn from our mistakes.
I think it is the businesses who think they can rely on as little people as possible, that are going to be in trouble once the cost of using AI are going to go up. At the moment the AI companies are running on big losses, and at one time they are going to generate money instead of spending it.
Even as someone that has experience I don't know everything. So are they going to fire me because I need a ramp up period to learn a new thing when asked?
When I did hiring interviews, I looked for intelligent questions when hiring a junior. Even if they were beside the point, it showed the curiosity to understand the problem.
It is a rough time now because a lot of things are still up in the air, and nobody can tell what is going to land.
Exactly
The way to do this exercise is to launch the agent, give it the code and the logs and ask it what went wrong.
That is an option, but you still need to be sure the refactor works. AI companies love nothing else than letting an agent keep executing commands until the logs are successful.
It might be possible the log errors don't contain enough context for the agent to act on.
I provided some better ways to get to the cause, instead of looking at the code.
I made some assumptions, because I haven't seen the actual code and logs, and I don't know what the agent does.
Wasn’t allowed to use the agent…. That’s the issue
Complete nonsense, if they can generate 2000 lines with an agent, you should be able to use at agent to debug it.
This post highlights a, for me, rather concerning development in the field: By replacing juniors with AI models and only hiring seniors, we'll run out of seniors eventually. If there are no juniors today, there are no seniors tomorrow and seniors want to retire at some point. Most companies that replaced juniors with AI models trade a sustainable industry development for short-term profit maximizing. Sadly, I do see why they're doing this, though. After all, shareholders are a thing.
Regarding universities: Should they teach for students to get entry level jobs? I.e. system forensics, orchestration and architectural judgement? Or should they focus on teaching a solid foundation in software engineering, math and theoretical computer science so the students are able to become senior eventually? In my opinion, if universities would focus on the first, they're actually playing into the problem, since the juniors will then lack the foundations of good software engineering.
The question is, how the industry can break out of this and ensure that enough juniors get the opportunities they need to become seniors. People with 10+ years of experience don't suffer from this. They're senior enough that even if they want to pick up a different technology, their ramp up time is short enough and their expertise in other techs large enough to justify the investment of hiring them. People with 2 years of experience or less? Too much of an investment.
Thank you for bringing this issue to light.
Thank you for sharing your thoughts, you’ve said it all😊
I’m so sorry, Maame! That interview didn’t test anything at all. Maybe the guy himself had no idea how to find that bug, so in a moment of desperation he did the interview 😄 Fingers crossed that you land a really cool job!
Thank you Sylwia! 😊🤞🏾
Chances are the recruiter didn’t have the solution either.
Whenever an interview didn’t go my way, I always made a point to ask why — the criteria, the reasoning… And in a case like this, I would have asked for the actual solution.
If they didn’t have one, well, I’d draw the obvious conclusions.
i am assuming this could have also been the case
I'm very much in agreement with your thoughts.
I'd like to add one thing:
1) The guy who interviewed you is an idiot. There's no doubt about that.
2) Software always belongs to those who create it.
Even before AI, if a company lost its lead programmer, it was at serious risk.
I don't know how much it benefits a company to have all its code "created" by an entity with as same interests as a stone. I mean, the day there's a serious problem, that guy is screwed!
Rightly said
I agree with you priolo.
Are you freelancer?
I’m currently a part-time Computer Science student , but I do freelance part-time to sharpen my skills. I’m actively looking for internship opportunities at the moment, happy to chat if you have something in mind!
Wonderful to know you are exploring part time freelance to sharpen your skills, way to go.
I've been there too, struggling to find my footing in the industry. It's heartbreaking to hear that you were rejected for a junior role because of that stupid exercise. Universities need to start teaching what is actually in practice, not just how to write code for a specific job.
It is quite sad because universities these days even teach outmoded tech, it’s all about getting a good grade in most universities now
Universities have and always will teach "outdated tech". But what you can and should gain from learning that older technology or programming language is the foundational fundamentals that will allow you to adapt to what is used in the "modern tech stack".
Don't look at the older technology as a hinderance. Look at it and use it as the foundations of what all modern tech stacks have been built on top of.
You’re absolutely right about fundamentals, they do matter, and they’re what make adaptation possible in the long run.
Where I think the tension comes in is that universities are optimized for grades and theory, while the industry increasingly expects you to already know how to solve problems you’ve never encountered before. Bridging that gap isn’t trivial, especially when you’re expected to deeply understand the basics and somehow simulate years of real-world experience at the same time.
There are only so many hours in a day, and for students it often feels like you’re constantly behind no matter which direction you lean. You look around and see people who’ve already “figured it out” (the DSA gods and FAANG lovers) and it’s hard not to question whether you’re even cut out for this.
I also think mentorship is a big missing piece. In tech especially, good mentorship feels rare... not because people are unwilling, but often because everyone is stretched thin or still figuring things out themselves. Without that guidance, connecting theory to practice becomes much harder than it needs to be.
I don’t see foundations and real-world readiness as opposites, but right now, the burden of connecting them falls almost entirely on the individual, and that’s where a lot of people struggle.
Thanks for the advice Ken
Universities must pivot and rethink - if not, they'll go the way of the dinosaur ... if I'd be in the management of a university, or even just teaching at one, I'd be scrambling to rethink Uni's place in this brave new world, and how they can reposition ...
Reading code has always been more important than writing it, so nothing changed there. Don't lose hope, you will find a place, and no LLMs do not write perfect code and are pretty much useless without someone in the driver's seat. Keep your chin up, keep learning, and you will make it for sure.
I totally agree with your point. Keeping hope and courage is the best way to beat everyone in the domain :))
A 100%
Thank you!
Thank you!
So true
Kinda intimidating, edgy... but I guess it happens.
Please, don't give up.
Not giving up will just do better. 🙏🏿
Finding and fixing a bug hidden in application code was a requirement to pass entry level Vue.js exam few years ago, before AI agents even exist. It is a very useful skill and it always was.
The thing is schools almost never keep up
with the latest reality, because their plans are being composed by people who reflect their past experience and not the current reality. Especially now when things get obsolete within months.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.