Is Generative AI Replacing Junior Developers? Jalasoft Leaders Weigh In

Hey guys, thanks for the throw. We're here with Juan Salinas, who's vice president of business Development for Yof, and he's joined by Rolando Lora, who's a staff engineer and software engineer who is also leading up their agentic AI push. And we're talking about the impact that AI is gonna have on the need for junior developers.

Jp, welcome to the show. Hi, Mike. Thank you for having us on.

All right, Rolando, you too. How are you? I'm good, thank you, Manik.

All right, jp, let's start with you because there is this ongoing debate about, um, well, just how much are we gonna need junior developers going forward? Because the old guard will say that I don't have to assign out tasks to junior developers. I'm just going to give that to some sort of agent ai.

And I also know that you're the CEO of your company's university program, so you see some of this firsthand, but what's going on? Yeah, I mean, we have to confront this, uh, this truth now that junior developers are gonna soon become obsolete. And, and we understand it.

I mean, uh, coming from a 20 year, um, uh, company in the industry, we, we understand the significant impact that AI is gonna have in, in junior developers. It has always been difficult for junior developers to get into their, uh, first jobs in the high tech industry, but now it's gonna become even, even worse. Um, it was mainly difficult for them because there's a huge gap between the academy and the industry in the educational models.

So, uh, engineers graduating from college were maybe like 20, 30% job ready when they graduated. And now the same agents that they're using to help them in their assignments in, in the university, these agents are gonna compete for their first, uh, jobs as well. So if you think of it, um, when you get a junior developer in your team, you give them tasks, I mean, clear instructions, low risk tasks, uh, things that they can start, I mean, uh, working on without creating, uh, much impact.

And that's exactly what, right now people are dedicating to AI agents. So that's gonna be a tough situation for junior developers. Now, Rolando, you are neither too old, nor too young, so you're kind of in the middle of this conversation, but you're working on these projects.

But I will talk to some junior developers and they'll say the exact opposite. They'll say, this stuff is great. It enables me as a young developer to do all kinds of things that I probably wouldn't have been able to do on my own without some sort of specialist.

And basically they're saying, Hey, boomer, move over. Yes. Um, absolutely, Mike.

Um, well, I'm, um, we were having like, uh, a lot of experience, um, with the real teams, uh, that are using artificial intelligence. And, uh, we can see that a lot of the buzz, for example, around vibe coding, um, it's, uh, real, uh, it is a time when you can actually, uh, do a lot of stuff without, uh, uh, the need of grasping, uh, language skills and, uh, learning how to code. And, um, we, for example, in larger projects, sometimes legacy projects, we use a lot of five coding, like, uh, product owners and, uh, people that are not exactly, developers are, are using it, uh, to communicate ideas because they can prototype something, um, that maybe does not work at a hundred percent, but they can communicate an idea.

But, um, in our case, uh, I'm looking at senior developers that are using and, uh, not necessarily five coding, but agent ai. And, um, they are able to grasp like, uh, really complex tasks and features, but, uh, the way that they are using the AI is completely different. Uh, they, they spend like, um, most of the time reviewing code, but, uh, they are able to generate, uh, different kind of solutions per day.

So it is a tool that allow them to explore the solution space and, uh, pick up the best solution. So in, in that case, for senior developers, we are not looking like at, uh, productivity gains in terms of, uh, how much they can do per day. But, uh, there is an improving quality because they can deliver maybe the same feature, one feature per day, but, um, with a lot more quality because they are able to prototype a lot of ideas and solutions.

And, um, in order for junior developers to do that, um, they, they do need to have a bit more of experience, and they need to have the skills. So that's what we are trying to figure out how we can, um, teach those skills to, to the students. Jp, do you think that maybe there's room for junior developers in this sense, to Rolando's point, we're gonna see a lot of vibe coding, or that may not be done by professional developers or people with any kind of developer training, so I don't think the senior developers want to clean up that mess.

So maybe the junior developers can go in and clean up that mess. Clean. Yeah, potentially.

But I think we're taking it from a different angle in, I mean, this is an opportunity for us to really try to graduate mid-level developers from college. So the junior tag is just gonna be removed completely if we train them well in the university, I mean, they will graduate, uh, as mid-level developers with the maturity and experience and the software development lifecycle exposure, right? In college.

So, and they're gonna be exposed to AI to use it effectively, efficiently, and responsibly and safely for their teams. So that's kind of our approach. I mean, okay, we're junior developers are gonna become obsolete.

Okay, we take it, but we're gonna graduate mid-level developers then right from college. So, Rolando, going back in time to when you were a student, um, what did you learn in college and then when you came into work, how big a gap was that between when you could actually be useful for the company that was hiring you? There was actually a, a big gap.

Uh, I, I spent a lot of time programming in college, and, uh, my thesis also, um, was related to a problem that was really difficult to solve. So when, when I get to the industry, and actually my first job was, uh, jealous of, so I, I was like, um, very confident that, uh, my skills were going going to be enough for, for the job. But, um, when I started to deal with real world problems, that's what where my perspective, uh, shifted a bit because, uh, and, and I think that is happening now because, uh, people, uh, is, um, judging all these AI tools mostly on, on a small or, or hobby projects, and they perform extremely well.

But, um, if you look at the, at a company that has a legacy project that has huge code base and, uh, very complex infrastructure and deployment, um, actually these tools are, are not able to perform that well in that kind of environment. And I, I think the same happened to me when, when I came for my first job, that, uh, it took me a while to understand everything and, uh, and make sure that, for example, uh, a small fix that I could enter as a junior developer did not end up, uh, breaking something critical. So it was really difficult, but I think that now it is a lot easier for, for this new junior developer to join a company that is working on large projects because a, actually these tools are very good for that.

They can help you understand a large code base and, and the inception and the bootstrap, uh, time in a project, I think it's much, much faster. So it, it's a, a great opportunity that we have, um, to actually, uh, change the, the role of the junior developer a little bit and, uh, and make, make them fit in, in this kind of projects. They, they will perform much better than we did at that time.

So jp, how does the university program need to change then to turn out more mid-level developers versus junior developers? That's a great question. We take this, this approach, we have a PBL, which is very common project-based learning, but the problem we're seeing is who's designing these problems?

When we get interns from throughout South America or even the us, we get senior students that never work in, uh, in a team, really with a common code repository or using methodologies. So they, they work in several projects, but on their own. So we also take the approach of PBL, uh, from the start, but we have a close connection with the industry thanks to JA of, so really we really mimic, uh, real case scenarios and challenges.

We create these projects for them that will expose them to the complete software development life cycle incrementally throughout the terms in the program. But they will get to experience the whole software development life cycle that is learning to work in teams communication with which is overlooked, uh, uh, fairly enough in, in the, in, in the education industry, um, methodologies, problem solving, judgment, um, starting to create more challenges pro uh, projects as they go as well. And having active engineers teaching them on how to resolve this problem, guiding them on how to resolve this problem is, is key.

Our faculty model has professors, uh, of record that have their masters and years of years of experience in the academy, but it's also mandatory to have faculty practitioners, which are active engineers in the industry, given their time to train the future, uh, developers as well. So I think this point for us was really key. And also the curricula was designed by, uh, by an architect as a software architect, uh, closely monitored by our chief academic officers as, uh, as well.

But I mean, this combination of academy and industry throughout the program, I think is key. Rolando, what is the future of application development in the age of AI agents? And I'm asking the question because some folks say, we're gonna have this small army of AI agents that are just standing by, and every time I want to do something, one of them will pop up and say, pick me, pick me, pick me.

And other folks are saying, I'm gonna have some sort of master, super agent that manages all the other agents, and it's gonna be more like the head butler and I might even give this thing a name. What is this gonna look like? Yes.

Uh, I, I think it, it'll be a combination of both. But, um, what I'm looking at is that, um, there has been like, like a lot of talk about prompt engineering, but um, there is a term that, uh, it, it, it's gaining traction that is, uh, context engineering. And, uh, I think that, uh, senior developers are starting to work that way.

And that is basically, um, try to get an agent or, or a set of agents to, to solve a difficult problem in one shot. So in instead of iterating through instructions and, and, and trying to get to a goal, um, through many iterations, they are, uh, trying to think in terms of, uh, what is the best context I can give to my agent or my agent to solve this problem? And, uh, this is very related to, uh, systems thinking and, uh, and, and abstraction in a way that, uh, future engineers will, uh, abstract a lot of the information that needs to be needed by the agent to solve the problem.

And they will, uh, let the agents work. Uh, in our case, uh, we are usually working with, uh, one agent or maybe two, or maybe conducting like deep research on other topics, uh, that we also send agents to do deep research. And it's very useful.

So in the, in the future, you, you will, you will have that, but the, there, there will be like other use cases that, uh, maybe will be more, um, targeted towards non-technical users that are trying to, um, promote an idea like innovators that want to release an idea to the market. And maybe that idea is not that complex in terms of, um, implementation. So these, uh, creative people, innovators will, will probably have like one master agent that will take care of everything and come back with the result.

So I, I, I can see both of the things happening. Jp, last question. What's your best advice to developers and the people who are trying to hire him these days?

'cause nobody's quite sure what the future look like. Uh, sure. I mean, uh, there's obviously, um, gonna be room for developers for, for a long time, so don't, don't get scared about AI and, and losing, I mean, um, the, the, this profession in, in the short term, I mean, that's, uh, not gonna happen.

I mean, they need to start educating themselves of our AI and really understanding AI's capabilities. I mean, there's a lot of buzz around, um, productivity gains like, or, or Orlando was saying, but is it really that much of an impact in the productivity or is it more like in the quality of the code that you're gonna be able to produce thanks to AI as well? And there's different levels of AI that you can be exposed to, like augmentation, collaborating with AI in resolving problems, understanding problems and concepts, discussing about solutions, and then automation, delegating pieces of tasks to these, uh, agents like call reviews, debugging, creating the release notes and things that take time from you.

And then lastly, when could you really use agency, which is when you delegate entirely a whole prototype to the agent. So it really depends on what task you are, um, confronting. And, uh, you need to be smart on how you're gonna use the different levels of ai, uh, to, to do your, to do your job.

Yeah. All right, folks. Well, the one thing that is for certain is there's no going back.

So mean, might as well all go forward together and see what happens next. Exactly. Yes.

Great. Hey guys, thanks for being on the show. All right, thank you.

What a pleasure. Back to you. Thank guys in the studio.