I am a subscribed paid user since the debut of ChatGPT.

It was fine throughout the year 2022 to 2023, helping me a lot in creative tasks and automation.

My nightmare started since around May 2024, which GPT-4o released while I just relocated to the UK.

The claims “coding is dead”, “software engineering is done”, “programmers are getting replaced” start spreading across the Internet like virus. Vibe coders claims the can create Apps with zero technical knowledge and selling courses everywhere.

My soul was tortured for a year, lost in confidence, doubting my decades experiences, nervous about my career and stressed of being redundant despite I start learning machine learning and data science since 2014 (Andrew Ng’s ML course on coursera and Microsoft courses certified), learned most cutting edge technologies like RAG, vector databases, MCP etc, due to the dreadful hype from all the AI bros.

The lifespan of Vibe coding hype is relatively short. Coined at February 2025 by Andrej Karpathy, and the author released a project with 8000 lines hand written code at mid October 2025.

After much thought and struggles over these two years, I finally get recovered and able to write this article. Consolidating what I have figured out.

In short: “We are not getting replaced, before I am getting retired.”

The feedback loop

Dunning-Kruger Effect and Vibe Coders

https://upload.wikimedia.org/wikipedia/commons/4/46/Dunning%E2%80%93Kruger_Effect_01.svg

Vibe coder get Software Engineering very wrong by thinking we just sit and type something to earn relatively high pay. Coding is not typing which is the work of typist. The value of inputting text keep diminishing for decades since the dawn of digital world. Software Engineering is about thinking and communication.

Programming itself is just a tiny tip of an iceberg, they think asking AI to copy some React code from others, running on their localhost then claiming they created a groundbreaking app is the end of story that will change them into millionaire over the night.

People paid a lawyer not because of the papers generated, the value not on typing the text but the content itself, the legal advice from a professional which the lawyer bet their name on being accountable and responsible. In Software Engineering context, our value is all because of one word: “Production”. When production incident occur, can you say you are not accountable because AI write the code?

Any commercial product will face problems like:

• Security & Privacy
• Scalability & Performance
• Availability & Reliability
• Compliance and regulations

Some people jealous others having a better life: They think musicians just finger moving around piano, producing random tones. They think illustrators and graphic designers just clicking mouse.
They think programmers are just typist. They ask “why all these people deserve better life than me?”

When AI launched, they think they get a gun finally, wiping all the needles in their eyes, claiming everyone are doomed, except them.

Smart people knows every professional:

• Got talent
• Keep learning
• Highly self-disciplined
• Admire others

In software engineering industry, it is well known one of the most harsh job in the world. Nothing change fast and massively like technology, not everyone can sustain the stress of hyper focus, good at logic and mathematics and keep sharping our tool belt every single day. I can say Software Engineer must be smart enough to capable day to day job.

These people just shows their self hate to their incompetence on the above.

Most importantly, they forget any professionals, either songwriter, illustrators and software engineers use AI too, probably use them better than the haters.

Quality will always persists, what’s matters

Tailor persist century after the debut of loom machine.

Doctor will exists even surgery can be done fully automatically by robotics.

Anything that is related to life, getting paid by using will need someone bet their name on to be responsible when thing happens.

Those self-haters will never able to wipe us because they never have the same competence as us to being able to be accountable.

When AI and machines can help speeding up our tasks, improving our productivity by doing 95% repetitive tasks that doesn’t matter the most, we have more time put our expertise focus on the last few percent hand-made customization that make a difference.

Example, fashion designer often cut and stich mass produced garments, clothes for their unique designs.

AI can help you get more works done (quantity), only you can help producing good works (quality).

2. The remaining few % is the differentiator of personal touch, taste and quality.

3. What vs How. What matters, not how. Taking refactoring as an example, it doesn’t matter whether you manually change a variable name line by line, use an IDE command, or use AI. What’s matter is what you refactored and why.

4. For professionals like illustrators, software engineers, writers, musicians, AI will only speed up on the noise and non-important things of our work, let us spend more time on what’s important: Creativity.

5. People who wanted cheap, massively produced goods are not your target audience even without AI.

The HOW doesn’t matter, it has no differences whether it is hand typed, IDE intellisense typed or Gen AI typed. The typing doesn’t has much value. The WHAT has.

Will software engineering done?

1. LLM cannot reach 100%, otherwise curve overfitting.
2. The remaining few % is the differentiator of personal touch, taste and quality.
3. Vibe coder / casual AI user cannot do [2] due to they don’t have the professional knowledge and training.
4. 0Code PM vs SWE, SWE can self manage small project with some PM skills or not needing a PM at all. PM cannot produce a production grade product by vibe coding. Nevertheless, PM is needed any for something scaled to an extent that a dedicated coordinator is needed.

Example of my AI assisted workflow

What AI best to do is when you know exactly what it need to do, being your automatic typing machine but you do the thinker. Using my refactoring case as an example, what matters is the refactoring I asked AI to type for me instead of opening hundreds of files and read the lines one by one.

Some people use AI to generate huge unmaintainable code / tech debt, while I used AI to dramatically reduce it.

• Telling AI to fix the nullability checks with specific patterns, with example.

Verdict

From Full-Stackoverflow Developer to GenAI Developer How could we survive under the tide as a software engineer There is a trend on social networks like Linkedin / Thread in last couple of years. Many claims how productive GenAI is and the death of software engineer will come soon. This article is my opinion based on personal experience. Speaking of GenAI we know OpenAI’s ChatGPT and GitHub Copilot. As a paid user since the debut of them both, it absolutely helped me a lot on swallow information retrieval. Saved time on browsing each ad filled webpage from Google search. Yes, GenAI lowered the barrier to enter the software development industry to almost zero. However it is not a silver bullet for all through. There are countless times I have to tell Copilot that its result is wrong or simply generating long pages of outdated useless code.  Some people says GenAI is another industrial revolution. Do you remember you may heard the same when 3D printing and blockchain comes out? What history tell us is new jobs and roles comes as new technology emerged, some jobs will have a shift in responsibility but not entirely vanish into dust. Let’s take tailor as an example: Did sewing machine replaced tailors? Did industrial revolution and mass production replaced tailoring? Not really, although a large amount of them have their target audience changed. If you think code like documents, they share some characteristics with clothes as well. Standardised, mass produced clothes are affordable and suits most people’s needs. This also applies to program code. With GenAI, the cost to generate a piece of canned code is almost nothing, those can be used directly and suit very basic needs. But there are cases that those generated code does not fit. Sometimes we would think: “I want my T-shirt to have a stitched flower.” Since you do not know how to do it (and do it nicely) or don’t have the tools, you go find someone that can have the tool and knows how to operate it to get you what you need. Software Engineers, in such sense, is very similar.  What GenAI can do is to generate typical things for your MVP. You want to differentiate with your competitors, that little extra mile is key. Not to mention custom business logics differs between companies. Either you know how to fix or customise the code or you will need someone else. Remember software engineering involves both typing and thinking. Typing is repeative and cumbersome, and GenAI is removing this part. In the past we communicate with computers by typing code. The communication part does not change, just it is largely replaced by communication with your GenAI tool. The most valuable part of your skill and experience is the thinking portion because it involves creation, which tools are not good at (at the moment). In future, every line of code you actually written should be well crafted, like a tailored tuxedo. Let the tools do the heavy lifting.  You are software engineer, by engineering you achieve your goal with the help with tools. When civil engineers and mechanical engineers built bridge and air-crafts, they use tools, and better tools let them built more awesome things, so do we. If you understand the underlying technology of GenAI, you won’t afraid you will lost your job, unless you are a Full-GenAI-Developer. GenAI is a keyframe technology that you should know but at the same time you shouldn’t too worry about it or those Full-GenAI-Developer bloom from the void in last couple of years. The bloom of developers exists since web2.0 that video and learning platforms substainsally lowered the barrier before, GenAI just push this further.  GenAI works well on generating code, but still not good at setting come your infra or publishing your app. Still far from able to help you automate these things. The ability that can launch a product from zero to one, end to end will be crucial.

Why as you getting into more senior role, your focus should change from writing less code? This is because everyone got 24hrs a day, even you are a 10x you cannot do twenty people amount of work without burnout. Spent time on writing good documents, system blueprints, onboarding guide etc. No matter you’re in manager path or IC path, helping new hire to get up and running asap is more beneficial than putting all burden on your own shoulder and create silo.