AI is changing software engineering. But is it really the end?

Disclaimer: This is entirely my personal opinion, based on my 12 years of experience as a developer. It is not a research report or a guaranteed prediction. I could be wrong, and your situation may be different from mine. Read it, think about it, and form your own opinion.

I am scared too

I know what we are seeing right now makes it feel like everyone’s career is coming to an end. Honestly, all of us, including me, are worried about what is happening to the job market.

Am I afraid of becoming irrelevant? Yes, I am.

Let me explain why.

AI has changed two things for me.

First, if I was already 10x productive, AI has made me 100x more productive. Things I always wanted to build but couldn’t because of a lack of time, money or team members are now possible.

Second, the same things I can do are now possible for another 1,000 people.

And that’s where the fear begins.

Two different kinds of fear

There are actually two types of fear here.

The first is about my own output. As a software engineer, I can now build things so quickly that even my own output surprises me. Previously, a feature might require 4–5 people across backend, frontend and QA. Today, I can build much of it myself. That means fewer people are needed for the same amount of work.

The second fear is about everyone else. We shouldn’t only think about developers. There are many other groups of people: business owners, doctors, lawyers, students, freelancers and many more. They all have access to the same tools now.

Let’s keep them aside for a moment. I’ll come back to them later.

A thought experiment: what if everyone became a skilled software engineer?

Let’s imagine that everyone in the world suddenly gets the same skills I have developed over the last 12 years. Everyone becomes a skilled software engineer, using nothing more than a $100 Claude subscription. Let’s also assume everyone produces the same quality of output.

What happens next?

Someone who needs an e-commerce website will build it, deploy it, maintain it and handle customer support themselves.
Even Jeff Bezos could build his own tools using the same subscription!
But think about it differently. Will Mukesh Ambani actually sit down and do all this work himself?

Or will a typical SaaS founder like me do it? Someone who is addicted to tokens, constantly spending money on them, almost like trying to fly without fuel, except that the fuel is human effort. And on top of that, we have to pay a heavy tax to Uncle Sam or Claude.

We already know the answer. Ambani won’t do it.

And that’s where the first crack appears in this entire fear story.

Everyone has the tools, but not everyone has the time

Let me explain something I strongly believe.

Even if everyone in the world gets a $100 AI subscription and becomes equally powerful, someone still needs to operate these tools.

The world has never worked in a way where every business owner has enough time to do everything themselves. Business owners won’t spend their nights burning tokens and building tools just because they can.

Think about it. Almost every computer has had Excel for years. Yet accountants still have jobs. WordPress has been free for more than a decade. Yet web agencies still exist.

Why? Because people don’t just pay for the ability to do something. They pay for time, attention and responsibility. They want someone else to take ownership and get the job done.

Now you might say, “Jeff Bezos may not build his own tools, but I cannot build something for him either.”

That’s where we come back to the basics: networking, and getting yourself into the best companies. These things mattered before AI. They matter today, and they will matter even more in the future.

Critical business software will never be 100% vibe-coded

Software developers don’t just serve small businesses. They also work for large enterprises and some of the biggest companies in the world.

Imagine the Dhaka Stock Exchange needs a new software system. Will its chairman simply sit down, vibe-code the entire system and approve it for production? Would anyone seriously approve that? Of course not.

Why? Because the question isn’t just about writing code. It’s about responsibility.

Who takes responsibility if something breaks?
Who makes sure the system passes an audit?
Who answers the regulators?

If one wrong transaction holds up thousands of crores, you cannot simply say, “AI wrote it.”

You might argue that these types of software are limited in number. I would argue that we don’t see enough of these opportunities because many of us are too busy building CMS, LMS and e-commerce products.

Just look at Bangladesh’s financial infrastructure. How many local tools are being used across banks, mobile financial services, insurance companies and capital markets? There is still a huge opportunity for local engineers in these core systems.

My point is simple: software that handles serious money, people’s lives and critical business operations will never be 100% vibe-coded.

You can use AI to generate every line of code or build every feature. That’s fine. But you still need testing, real-world data validation, security reviews, and someone who understands the entire system.

Three market segments from a business perspective

There is another market segment we all know. Think about a senior from your neighbourhood, a friend of your father, or someone you know who runs a business. They might need:

A hotel listing website.
An e-commerce website.
An ERP system.
A CMS.
A portfolio website.
A mobile app for product listings and checkout.

Honestly, this type of basic work has been available for years. You could find many of these products on CodeCanyon or even get free alternatives.

But who will build something custom for a business? Will the business owner spend two hours every day, or three to four months, learning and building it themselves? Or will they hire someone?

I believe most will hire someone.

From a business perspective, I would divide this market into three segments.

High-level business leaders. These people have serious software requirements. They are not going to build everything themselves. They have teams, responsibilities and businesses to run. They need people to get the work done.
Serious businesses. Think about aviation, healthcare, medicine, fintech and banking. Money and human lives are involved. AI will be part of these businesses, but engineers will still be needed to take responsibility.
A huge market of smaller businesses. These businesses might only need an e-commerce page or a small internal tool. Some tech-savvy doctors or lawyers will build their own tools. But most people will still buy ready-made, AI-powered products because they don’t have the time.

Thin-wrapper SaaS products will face serious pressure. Even larger companies like Zoho and Monday may face new challenges. Low-end gigs on Upwork and Fiverr will probably take the biggest hit.

Let’s summarize this round with three points:

Everyone is empowered, but many people don’t have the time to build things themselves.
Some people will build their own tools.
Some products genuinely need detailed attention. You can write 100% of the code with AI, but you still need testing, data validation, scalability and safety.

This is teamwork. The team doesn’t need to be as big as before. A team of 4–5 might do what previously required 10 people.

The market is moving towards more serious products

Previously, we might spend two years building a startup. Today, for a CMS-type product, we can go from an idea to an MVP and initial market validation in just two to three weeks.

But one old startup rule hasn’t changed: 90 out of 100 startups will still fail. The difference is that we can now make many more attempts. So there will be much more noise in the market.

The companies that survive will hire people. What kind of people will they hire, and why? This is where business products, QA, DevOps and AI-powered development all come together.

I look at AI as a multiplier. Think of your existing ability as N. AI increases your output based on that ability. The bigger your N, the more you can get out of AI.

This is also why AI is making education more challenging. People starting with a smaller foundation may find it increasingly difficult to catch up.

AI is like a very powerful sword, a gun, or even a genie. You can ask it to do almost anything. But the sword is still in someone’s hands. And that person matters.

Now, let’s get into more detail.

Will everyone buy a $100 subscription? No. Where will people who don’t buy these subscriptions get their automation and software? From people like you and me.

So what will companies do? Will they hire a separate person for every position? No. They might use one $100 subscription and a team of three or four developers to build much larger projects.

Remember when Facebook acquired Instagram? Instagram had only around 13 employees at the time. WhatsApp had around 50. Small teams building big products is nothing new. AI is simply making it more normal.

What will companies need now?

Not just CRUD developers, but an FDE mindset. FDE means Forward Deployed Engineer: someone who works closely with the client, understands their actual problems and builds solutions around them. Even if you’re a coder, you need this mindset.
Engineers who think like product owners. Someone who understands what needs to be built, why it needs to be built, and what should not be built.
Combined DevOps, testing and software engineering skills. Companies may not need completely separate people for every task. They will value engineers who can handle multiple parts of the development process.
Polyglot engineers. Companies won’t only look for C++ or Python developers. Language barriers are becoming less important, and basic CRUD coding is becoming easier to automate. You need to be able to work with different stacks and get things done.

The idea is simple. Someone with a $100 subscription will take responsibility for running a business throughout the day. As software developers, you and I have access to tools that everyone else can also use. Some people will use them as a hobby or to solve their own problems. But you will use them to serve people who don’t have access to a $100 subscription, or even a $20 one. And even if they have access, they may not have the time.

Now comes the important part: not everyone is equal

Until now, I deliberately made one assumption: everyone with a $100 subscription has the same skills as a software engineer. In reality, the opposite is true.

A software engineer with 10 years of experience and a $100 subscription is far ahead of someone with no experience.
Even a junior developer with a $100 subscription is far ahead of someone who doesn’t understand software engineering at all.

Why? Remember the idea of AI being a multiplier and your existing ability being N? The bigger your N, the bigger the gap.

AI can bring beginners closer to experienced developers when it comes to small and simple tasks. But as systems become more complex, experience becomes increasingly important.

Let’s look at a simple example. Assume an illustrative model where output with AI = N².

No experience (N = 1): without AI = 1 → with AI = 1
Junior developer (N = 3): without AI = 3 → with AI = 9
10-year experienced SWE (N = 10): without AI = 10 → with AI = 100

Without AI, the difference between an experienced engineer and someone with no experience is 10x. With AI, that difference becomes 100x in this simplified example.

One more thing I want to make clear. Low-cost, basic web development work is disappearing. But much of this market was already declining before AI. We just noticed it later. Nobody wants to invest heavily in another basic CRUD e-commerce application.

The startup game has changed.

Here’s a challenge

Ask someone to build a small application similar to Pathao’s driver-finding system. Give them a $100 AI subscription. It must include all the details: admin panel, security, scalability and product requirements.

If they can do it, I’ll give them $1,000. There are two conditions:

The system must match Pathao’s scale and requirements.
The total cost must be equal to or lower than Pathao’s cost. Not even one extra penny.

And this is only for the driver-finding system. Or think about bKash. Or an airline ticket booking system.

At this scale, a huge part of the cost isn’t writing code. It’s infrastructure, operations, thousands of edge cases, and handling incidents at 3 AM.

This is where people start making unrealistic claims. Someone builds a basic backend, UI and prototype of something like bKash and says, “That’s it! Everything is over!”

Yes, everything might be over at your level.

Someone else might build something that looks like Amazon, but they cannot go beyond the limits of what they themselves understand. No matter how expensive your AI subscription is, the quality of your thinking and your ability to solve problems directly affect what AI can produce.

And here’s another challenge to think about. You built the app using AI. Now who will handle marketing and sales? I’ll leave that question here and come back to it later.

Three layers of the job market

Now, you might tell me, “There are no jobs anymore.”

Yes, there may be no jobs where you’re currently looking. But I want you to divide the job market into three layers.

Layer 1: the jobs everyone sees

This is the market everyone talks about. And much of it is already changing.

Selling basic software services in this segment is becoming extremely difficult, at least within the networks we have access to. Whether it’s BDJobs or Upwork, AI is automating many low-level tasks. Nobody wants to hire a developer just to build an MVP or a small utility tool anymore.

If you’re building tools yourself, you have to handle sales and marketing too. This is called indie hacking. But remember, working alone in this market doesn’t mean you can build the next bKash or Pathao.

Layer 2: building tools for companies

The second layer is where companies hire you to build their internal tools and systems.

Companies that only build plugins or simple MVPs may struggle. But businesses that offer software services as part of a larger sales and marketing solution can continue to survive and hire people. These companies might interview you for a MERN stack position, but they will expect you to do much more.

What will they build? Software for:

Banks and financial services.
Mobile financial services.
Healthcare and medicine.
Agriculture.
Garments.
Sports technology.
Advertising technology.
Media technology.

You need to identify companies in these industries and understand their requirements. Many Fortune 500 companies also operate in Bangladesh.

Will you get a job in these companies after preparing for two or three months? Probably not. You need to understand the business domain, and that takes time.

Layer 3: product companies

The third layer consists of product companies. These are the companies that could become the next bKash, Pathao or Uber, or companies that have already become FAANG/MAANG-level businesses.

Getting into these companies was difficult before AI. It is difficult now, and it will probably become even harder. You won’t get these jobs with just three or four months of interview preparation. But these are still some of the most promising opportunities.

Let’s recap

The market you see every day is not the entire market. There are many opportunities outside your immediate view.

Both the visible and less visible parts of the market are changing. You’re mostly seeing the impact on the segment you’re connected to. That’s why the fear feels so big.

Focus on the things that cannot be done easily with a $100 Claude subscription. And remember, you cannot develop those skills through a one- or two-month course.

Let’s imagine that one day Claude and ChatGPT become the only tools available in the world. Even then, basic human thinking will matter. And if that future ever arrives, everyone will be affected, not just software developers. Perhaps there won’t be much work left for anyone.

But we are not there yet. Until then, we have to keep playing the game.

So, what should you learn?

If you’re entering this industry or already working in it, understand one thing first: things are becoming riskier, and building a successful career may take more time.

Everyone says coding is no longer necessary. But then you go to an interview and they ask you to reverse a binary tree. So focus on both sides, although perhaps with less focus on writing code line by line than before.

When people say “coding is gone”, they mostly mean writing syntax is becoming easier. Engineering is not gone. The entire learning process has changed.

If you don’t even look at every line of code yourself anymore, why should someone else care about every line you write? People will look at the product and the problem it solves.

Everyone can build an MVP with a $100 subscription. You need to prove that you can build and manage larger systems and infrastructure.

Yes, not every company in Bangladesh needs these skills. Not every job requires them. But if a business doesn’t need complex systems, it may not need a traditional software engineer either. I’m not talking about those jobs here.

Skills you should learn

System thinking: system design, architecture, and understanding how large infrastructure works, breaks and stays reliable.
Cloud and databases: not just CRUD queries. Learn indexing, replication, and where infrastructure costs come from.
AI engineering: learn how to use LLMs in real products, including evaluation, guardrails and cost control.
CI/CD, testing and observability: the more code AI writes, the more important testing and monitoring become.
A business domain: choose one area, such as fintech, healthcare, agriculture or garments, and understand it deeply. It is much harder to replace an engineer who understands the business.
Sales and marketing: if you cannot sell your skills, surviving will become difficult. Why do I write posts and make videos? Because when people know me, finding work becomes easier.

What should students and freshers do starting tomorrow?

Choose a real problem that exists in a nearby shop, clinic or business.
Use AI to build a solution, but take it all the way to production. Include deployment, monitoring, backups and security.
Get real people to use it. Ten real users are worth more than 1,000 imaginary users.
Share your entire journey publicly. Explain what broke, how you fixed it, and how much it cost. This is your new CV.
Don’t abandon DSA and computer science fundamentals. Interview doors still open through them.

What about people already working?

Use the time AI saves you to learn the next level of skills: system design, business domains and leadership.
If you’re stuck in a CRUD-heavy role, start planning your next move.
Build your network. Many opportunities in Layer 2 and Layer 3 come through people you know.

Three types of people: who will stay ahead?

Let me give you another perspective.

Getting a job by learning technical skills is becoming harder. But people who understand marketing and sales may be able to earn much more through their own services or products, without even having a traditional job.

How? Let’s look at three different profiles.

Profile A: the certificate collector

What they can do: completed a MERN course, built five tutorial projects, and have a GitHub account with a few basic apps.
After AI: AI can build most of what they know in 10 minutes. Their CV gets lost among thousands of others.
Where they stand: Layer 1, the most affected segment.

Profile B: the system builder

What they can do: understands a business domain, has experience running systems at scale, and knows how to handle production incidents.
After AI: AI makes them faster and increases their value.
Where they stand: Layer 2 and Layer 3.

Profile C: the builder-seller

What they can do: a good engineer who can also write, make videos and talk to clients.
After AI: uses AI to do the work of an entire small team and sells the solution directly.
Where they stand: creates their own opportunities through products and services.

The bad news for Profile A is that much of their skill set is now available inside a $100 subscription. The good news is that they can still move towards Profile B or C if they start today.

Profile B is relatively safe, but their income ceiling is often decided by their company. Profile C takes more risks, but has no fixed income ceiling.

In my opinion, the strongest combination is Profile B and C: someone who can build large systems and also explain to people why they need those systems.

Remember my earlier question? You built an app using AI. Who will handle marketing and sales?

The answer is right here: you will.

Alternative career paths for software engineers

If you cannot find a job, it doesn’t mean everything is over. Your software engineering skills can open at least six different career paths. Each has different risks and requires different amounts of time. But none of them is closed.

1. Indie hacking

Build your own product and sell it yourself. AI has made the cost of building software very low, so the real challenge is distribution. Who will find your product? Why will they buy it?

Who is it for? Profile C: someone who can both build and sell.
What do you need? A small niche, a payment gateway, marketing through SEO or content, and patience.
Risk: most products will not make money. Out of 10 products, perhaps only one or two will work. Start alongside your job or freelancing rather than depending on it immediately.

2. MLOps

Building a model is one thing. Running it in production is another. Deploying, versioning, monitoring, retraining models and managing GPU costs are all part of MLOps. As AI becomes part of more products, the need for these skills will increase.

Who is it for? People with strong backend and DevOps foundations.
What do you need? Docker, Kubernetes, CI/CD, model serving, monitoring and cloud GPUs.
Risk: there are still relatively few MLOps roles in Bangladesh. You may need to look at remote or international companies.

3. AI engineering

This is about putting LLMs into real products. RAG, agents, tool calling, evaluation, guardrails and cost control are all part of it. This is not necessarily research. It is engineering. And this is where a good backend developer can find opportunities quickly.

Who is it for? Someone who understands products and knows how to identify AI mistakes.
What do you need? API design, vector databases, prompt and evaluation pipelines, latency management and token cost optimization.
Risk: tools change every six months. Focus on fundamentals, not just frameworks.

4. Cloud engineering

The more software we build, the more infrastructure we need. Even AI-generated code needs somewhere to run. Cloud architecture, networking, security and cost optimization are skills that are not easily becoming commodities.

Who is it for? People who enjoy understanding how systems break and how to fix them.
What do you need? Deep knowledge of AWS, GCP or Azure. Also learn Terraform, Linux and networking. A certification can help.
Risk: certifications may help you get an interview, but real project experience helps you get the job.

5. Any business that uses technology

Many people don’t even consider this option. Importing products from China and selling them, selling agricultural products, or running almost any kind of business can give a software engineer a huge advantage. You can build your own website, inventory system, automation, advertising tools and analytics. Other business owners might have to pay someone else to do all of this.

Who is it for? Someone who understands business and isn’t afraid to get their hands dirty.
What do you need? Supply chain knowledge, cash flow management, customer service, and Facebook or Google Ads.
Risk: the biggest problems here are not related to code. They are product quality, delivery, and money getting stuck. Start small.

6. Build utility tools and sell them as services

Build small tools such as invoice generators, WhatsApp automation, reporting tools and data-cleaning scripts, and sell them as services to specific businesses.

Remember those people who don’t have access to a $100 or even a $20 subscription? Or those who simply don’t have the time? They can become your customers.

Who is it for? Someone who can talk to clients and understand their problems.
What do you need? Choose an industry, create a small monthly pricing model, and provide good support.
Risk: the tool itself can be copied easily. What keeps you in business is your relationship with customers and the quality of your service.

None of these six paths is a shortcut. But they all have one thing in common: you cannot remain just someone who writes code. You have to become someone who solves problems.

Final thoughts

The fear is real. I have it, and you probably have it too.

But the market you’re afraid of is not the entire market. It’s only the part you can see right now.

AI is making powerful capabilities available to everyone. But it is not giving everyone equal amounts of time, responsibility, judgment, or the ability to reach other people. And wherever these things matter, there will still be opportunities.

So don’t look for shortcuts. You cannot win this game with a two-month course.

Build your N. Understand a business domain. Learn how to build large systems. And make yourself known to people.

Which profile are you in today, and which one do you want to become? Write to me and tell me.

Essay