➡ AI is rapidly advancing in software engineering, potentially pushing some tech workers out of jobs. However, not all computer science jobs will disappear; some will thrive. The impact of AI varies across different tech roles, with junior coders being most affected. AI can now perform tasks like creating database schemas and data entry screens, tasks that would have taken human programmers several hours to complete.
➡ AI can replace some roles in computer science, like junior coders and basic database engineers, because it can perform tasks like coding and creating database schemas. However, it can’t replace more advanced roles, like systems analysts, data scientists, and architects, because these roles require creativity, problem-solving, and understanding of complex systems that AI currently can’t replicate. Jobs like system admins, operators, and cybersecurity experts are also safe because they involve physical tasks or dealing with new, unique problems. However, tech support and training roles could be replaced by AI. The article also criticizes colleges for focusing too much on teaching coding and not enough on theoretical knowledge and problem-solving skills.
➡ This text emphasizes the importance of understanding AI’s limitations and using it as a tool rather than relying on it completely. It suggests that students studying computer science should specialize in a specific area and learn ‘Vibe coding’, a more intuitive approach to programming. The text also highlights the value of creativity and human insight in developing new applications, which AI cannot replicate. Lastly, it encourages the development of hybrid knowledge and domain expertise, using the author’s experience in medical informatics as an example.

Many of you have heard of AI vibe coding. This is when you just tell AI what to do and AI writes the code in any programming language. And I’ll be frank with you, if there’s an area where AI is moving the fastest, it’s in software engineering. And unfortunately the AI developers are pushing a segment of tech people potentially out of the job market. Several studies show a huge drop in entry level jobs, 20 to 25% for programmers and with the exception of programmers in AI. Well folks, this is a topic which I’m particularly equipped to answer.

As a chief software architect of many enterprise systems, I’ve had to lead software development teams. I’ve also had a long career in software development in various roles, including building full enterprise systems by myself. But the particular thing that allows me to answer this question even more is that I do vibe coding myself, meaning I now also use AI to write code. The question though is if new college students should stop going to computer science and software engineering as a career choice. And then the related question is if you’re a professor or university administrator in the computer science department, what should you teach so that your students get jobs? I have some answers and it’s nuanced.

There’s no direct answer that computer science jobs are gone or not. Actually, some will be gone, some will thrive. It depends. Stay right there. I don’t really talk about my background much as this is a privacy channel and a lot of my past is covered by so many non disclosures that I couldn’t even talk about it even if I wanted to. I was offered a first job to start training as an assembly language programmer at some large company, but I never actually did that training. I got a few months in as a management trainee before someone realized I had some unique skills.

So I quickly got promoted to the position of systems analyst. My job was to interface with users and then create a system solution. I had to spec out changes and design the operation of screens on a mainframe system. I then passed the specs of the project to the programming manager who then assigned it as a project. I was also tasked with designing a Monte Carlo simulation model which was used to forecast business problems. I programmed this myself in FORTRAN on an IBM mainframe. Many career things have happened after this, including transitioning to becoming a management consultant.

But after many years I formed my own software development company. My first commercial product was a great graphics products similar to PowerPoint. Then I ventured into enterprise systems since I had a lot of experience with that from the past. I built accounting systems, legal systems, HR Systems and so on. Finally, I created medical software, initially just for billing and collection, and later on I created full medical software suites that ran hospitals and large clinics. These software suites handled everything in an enterprise, from accounting labs, radiology, full electronic health records, to external interfaces passing data to other systems.

At the very beginning it was just a solo programming operation. And then as I had bigger companies, then I had a programming team. Then lately in the modern computing era, I created the Braxme app solo after I left my project. Prior Career Jobs in IT in order to analyze the effects of AI in it, we have to break tech positions into several discrete jobs. When I was running a small company, most of the tech jobs were just done by me. But in a large enterprise, these would actually be several jobs. Architect the architect would have the big picture design of the application based on his analysis of the enterprise needs.

This would have been documented very clearly. The architect would determine the programming languages to be used, the hardware and network architecture, the database architecture, the security architecture, and also include a scalability plan. Since the design may change as the user base expands, this could be specific enough to even create screen designs for a general user interface. But the specifics in design will fall to the next role. UX UI Graphics Designer Sometimes this is one person, or could be two. In the modern app world that powers big tech, this was a monstrously important job. This is the type of person that created the design of a Facebook or YouTube or a Twitter, meaning these are the visible signs of modern technology made possible by touch screens and mobile hardware.

They understood things like visual branding, iconography, and most importantly, user behavior. There are very few people with the skill sets for these, and to be honest, this is never taught in school. And the experts are all in Silicon Valley and the same experts are involved with different apps. This is not one of my skill sets, so mostly I just copy what someone else did. The fascinating thing is that some people are able to figure out what’s intuitive in the behavior of apps. With young people and those born in the old PC world can’t relate. Database engineers now this is a separate specialty.

These are the people who designed the database. A lot of this is handled by the architect who has a big picture of what’s needed though parts of it are delegated. But in a large enterprise system, especially after a long lifetime, a lot of this is parceled out to database experts. These people would tweak the database for schema optimization, indexing advanced concepts like sharding or funneling of data to SQL versus NoSQL. Often they also manage store procedures, backups, replication and then the data analysis tools. Honestly, this has become so specialized now that in large enterprises like a Google, this has to be quite sophisticated programmers.

Now we get to the actual programmers and depending on the size of the company, this could be one person or many. Some may specialize in particular languages. For example, the enterprise systems I wrote were programmed in C and some of the staff made web based modules and concentrated on JavaScript, C, Sharp and HTML. Others created pieces in PHP. There has been quite a shift in developer composition lately. A lot of lower level programming jobs has been outsourced to countries like India. And this is often the big mistake that causes failures in projects. This is because some idea guy businessman wants to create an app for cheap but without the knowledge to do system architecture, UI design or database design.

Then he passes the entire project to some Indian programmer who has no idea what the specs are that are required by the user. This then causes the biggest roadblock to completing projects when the business people don’t know the elements that go in designing an application. Data scientists. This is the area that’s being taken advantage of by companies like Palantir. This is the data analysis side. These are specialists who understand the big picture of data and they just see the relationships of data. They then create brand new designs of user interfaces that allow users to view the data analysis in a different way as it is used by the military or intelligence agencies.

For example, this is another subspecialty that very few actually know about. Plus there’s a whole application side to creating this user interface for analysis that is not connected to the usual UI of a regular app. Ancillary support jobs. Then there’s the people that manage all these to make sure the hardware is always running and performing at the peak capacity. These include sysadmins, operators, tech support, CyberSecurity trainers and QA. Many of these folks are the ones with expertise in networking hardware devices and virtual machines, cybersecurity and data center management tools. There’s also the tech support crew that interacts with end users and maybe even manage the individual computers of those users.

Then there are also the QA testers that work closely with the developers, though these don’t necessarily come from the ranks of computer science people. Some of these roles may be handled by the data center handling the application. So not every company may require separate staff for these. Can AI replace all these jobs? Now that I broke it down into specific roles, we can really pinpoint how AI would become integrated into these roles, either as a replacement or As a co worker programmer, the most affected job in this group is the junior coder or junior programmer. And unfortunately this also removes the need for the outsourced programmer, often from India.

When you give a task to a junior coder, you give them the specs of what you want to do and then the coder gives you the finish code and response. I’ll give you a quick example of this. I asked the AI to write me a data entry screen for an HR demographics application. And first I asked AI to create the database schema. And then as the first application I asked for the demographic entry screen. I then asked it to make a general login screen. This kind of method of asking the AI to do the programming is called Vibe coding.

Now I will tell you that I’ve designed HR systems in the past. So with a few commands that created in a few seconds what would have taken me several hours to figure out. And not only that, I would have been bored to tears. The problem is that to give the specs to a junior programmer, I would have had to already provide the database design and even the exact fields I need on the first screen. A junior programmer will not figure it out on his own, but the AI did. So in this, the AI took part of the architect role, while AI Vibe coding took over the entire role of the junior coder.

So the initial result of this is bad news for junior coders, which are typically your entry level college graduate of computer science. I didn’t mention advanced programmers here. And the reason is that it is already difficult to even state a problem to the AI when the problem is very advanced. So those in advanced programming positions will not be affected by the AI anytime soon. While a junior coder could use Vibe coding without some domain expertise, it would still require checking by a more senior programmer database engineer. Now, as I already explained in the prior example, the AI created a database schema for me in a few seconds.

So there is no doubt that that an AI could also perform a lot of operations needing a basic database engineer like one trained in SQL. Unfortunately, we have the same problem here. These levels of these jobs can easily be replaced by AI Vibe coding. However, as you get into more complex optimization problems, I think AI will have a more difficult time. I’ve had performance issues in large databases and it is not clear from the SQL queries what the cause of the problem is. In large complex tables with complex relationships, these slowdowns can be caused by improper indexes, but these are not obvious.

Without examining the actual data, the AI will not see that data so it will guess. It will likely guess wrong. So advanced database engineers will not be easily replaced. Architect Systems analysts data scientists this group will literally be irreplaceable by AI for a very long time. And the reason is that developing enterprise systems is not solely an intellectual operation in one person’s brain. A real enterprise system means interviewing people to establish the needs of the enterprise. And this holistic view of the problem leads to a general systems design. The architect may have the big picture view of this.

The systems analyst is problem solving this with end users. And the data scientist is doing creative research by actually looking for patterns in the data. And so it cannot be done without the data. These people are not only involved in communication with various parts of an enterprise, they are by necessity creative people. So unfortunately, the abilities involved, including social interaction, are beyond the skill levels of AI. Currently, just to be clear, a systems analyst has to draw out responses from users to get a clear idea of what they want. In my personal experience, users don’t necessarily know what they want, they just want something to work, while an AI isn’t going to figure out that subtlety.

Also, an architect doesn’t take the original problem at face value. The entire role of the architect is to fashion a solution, often unique and with a focus that is best made to solve the problem. We all know that there are good software and bad software, and this results from good or bad. So software architects currently LLM AI unfortunately only knows what it has seen before. It is not able to fashion a unique solution, so it is limited in its abilities to replace this role. Also, the biggest limitation of AI is context. When you look at the enterprise systems.

I’ve developed some with many million lines of code that is beyond the capability of AI systems to assimilate. So you’re forced to deal with AI in small chunks. Thus it always misses the big picture. A software architect always remembers the big picture. But architects benefit well from vibe coding because it would be an excellent prototyping tool. In this sense, AI actually makes app development faster. System admins and operators Some of the roles of the system admins and operators are physical. Physical wiring, physical moving and plugging in of bare metal, configuring network interfaces by hand, and so on.

I guess it’s like jobs like plumbing and carpentry not likely to be replaced. I think AI tools will make these positions more efficient. However, the complexity of the jobs increases, as it always does. So these jobs are very safe. This is especially true because of the expansion of data centers. These kind of jobs are typically powered by certifications rather than specifically by a computer science degree though. Cybersecurity A lot of the mechanical aspects of cybersecurity will in fact be figured out by AI, and it is possible for AI to do bulk testing using a variety of pre made tools.

However, just like in systems architecture, a security researcher is not working from a list of known problems. Some are working on novel attacks that have never been seen before. AI can be quite a help here. It can speed up the process of recognizing current known threats. Cybersecurity is already getting more complex every day as it is hard to keep track of what’s new. AI can handle that, making sure you don’t miss anything, but new ideas and new approaches to attacks and solutions will come from humans. A basic level cyber security person may be at risk here, but I honestly don’t think AI is really used well for this yet anyway, so there’s time to move up the ranks.

What a cyber security person can do is use AI to do vibe coding of intrusion attacks. This will speed up the research. For example example if you know basically what you want to do, like write some attack code in python, I’m sure the AI can do that faster. But still the human is the one to dictate the vibe coding. So this career choice is likely safe Tech Support and Training unfortunately, this is a career choice that can be mostly replaced by AI, except for the edge cases. If you work for a small company with very specialized software, this may be okay for a while, but in long run this can be replaced, leaving only the really advanced experts Where Colleges are Failing Us A lot of college students and fresh college graduates are asking me about career choices and what advice I’d like to give.

But the main problem really is that the schools are failing us by teaching the wrong things. The least important thing is proficiency in raw coding. This is actually funny to me as I’m a particularly fast coder. Even in C programming, I could do 10 times the code of the nearest programmer in my team. But yet I’m telling you that being an expert in the language is not the most important and unfortunately this is the focus of the schools. If you’re an expert in some particular language, whatever the trendy thing is now, like Rust, then you’re supposedly easily hired.

I typically learn a new programming language in a day or two. It’s not that big of a deal since they’re all structurally the same. I can’t code fast in a new language since I may not remember the syntax and a lot of speed comes from memory. But let Me tell you, there should have been lots and lots of time spent more on theory. For example, future systems analysts should be taught that the intricacies of this simple chart, input, process, output, feedback, very basic back in the old days for me, but I’m not sure this is taught today.

Today it’s let’s start out with Python and program some app. Do young computer science students even do flowcharts still? Do they understand the algorithms involved? What about the secrets of the experts in Silicon Valley? What is the theoretical basis of modern UI design that makes the these designs work? I don’t know what the theory is, to be honest, because it is tied to understanding user behavior. What about presenting problems and theoretically coming up with algorithms or even generalized applications to try to solve them? The typical college focuses on students making fancy quick apps in little hackathons.

And a winning concept app will likely never make it in the real world where it is tested with reactions from real people. But yet it is the reactions of real people that is the basis for a theoretical understanding of a computing solution. And people understand this. Because AI cannot create something out of nothing, it is ill equipped to come up with an answer. This is something that some student who is taught concepts can figure out in hindsight. I did have some of this theoretical stuff in college because in the old days the actual programming was expensive, it was done on mainframe, so by necessity a lot of things had to focus on the theoretical, which amazingly worked well for me.

But now colleges are failing here. They need to revamp the curriculum here to focus less on creating coders, but focus more on systems analysts, architects, ui, ux, data science, AI and the theoretical basis of each. Folks, the mechanics of doing each are secondary to creating the theoretical knowledge for problem solving. In fact, I encourage schools to teach these students how to do vibe coding. Why waste time? Use AI like a calculator? They need to understand how to debug code, but the mechanics are not important. In C, you can waste an afternoon because you missed a semicolon.

That’s nonsense today and a waste of time where AI is weak. If you understand where AI is weak, then that’s the opportunity to find careers that cannot be eliminated. Generally, from what I’ve said above, AI is limited in context, which is its memory space. It’s pretty hard to give it even a few pages of code to analyze. Can it handle a megabyte of code? Probably not, at least not today. This means it can’t really fully design things with some overall big picture already in place. So it becomes an assembly of modules. But the problem here is that it takes some pretty advanced skills to see the flaws from module to module.

I encountered this a lot. For example, I would build a radiology system and it would not be a compatible structure to, let’s say, internal medicine. As I also said, AI is not good at creating novel ideas. It can copy, it can emulate. So when a problem needs a creative solution, it will fail. A lot of the killer apps developed over the years would not have been creatable by AI. I remember the killer apps in the past like VisiCalc and Lotus123, basically spreadsheets. No big deal now. But there was no such thing before. Only humans could have come up with that.

Only humans could have realized why apps like Facebook and Twitter became popular in the early days. Only humans could have realized how YouTube recommendations can power the platform, ironically using AI. So nothing has changed really. Finding the killer app or becoming an inventor is still unsurpassed as a way to gain notoriety. AI just allows you to prototype these faster career choices for young people. In it, the basic concept I just laid out really implies that people who study computer science have to really subspecialize. Pick an area that really interests you and you really need to go beyond the basics.

And schools need to allow students to achieve this subspecialization conceptually understand what I just said is the conceptual skills that that are more important than sheer knowledge of code. Good knowledge of one programming language is good. Do you need to know 10 languages? That’s purely a coder’s realm and you will not learn how to interface with human beings with this knowledge. Instead, go ahead and learn Vibe coding. Don’t get stressed about it. If you can vibe code like a mid level programmer, then you’re a mid level programmer, not a beginner. Prove it with some projects on GitHub and you’re good to go.

You will have a better chance of getting hired. Yes, practice building full applications using Vibe coding. Also, if you subspecialize in AI coding, that field is growing for now, so this doom and gloom is not global. Finally, a general point that you should know. I do not have a software engineering degree. I started off with an economics degree, though I did take programming courses and later I studied medicine. So believe it or not, hybrid knowledge is important because you could build something called domain expertise. In my case it became a field called medical informatics. That’s how I built the knowledge to create health information systems.

If you yourself don’t need this advice, I would venture guess that you know some young person that could use this advice, so pass it on. I realize it’s not about hardcore privacy, but it is indirectly because it is an AI topic. You see, you need to get to a point where you master AI before it masters you. Folks. Privacy is of course the main focus of this channel and I teach you technology so you understand the risk technology adds to your life. We have people who discuss these issues, including AI at my platform Brax May to support this channel.

We have some products in our store that provide the toolkit to retain privacy. We have braxmail, an email service with unlimited aliases and identity protection Brax Virtual phone Anonymous phone numbers bytesVPN for anonymizing your IP address. The Google phones phones free from big tech tracking the Brax 3 phone is on its second batch and is open for pre order right now@braxstech.net the first batch sold out shortly after release. Big thanks to everyone supporting us on patreon, locals and YouTube membership. You keep this channel alive. See you next time.
[tr:tra].

See more of Rob Braxman Tech on their Public Channel and the MPN Rob Braxman Tech channel.