I often wonder: with AI’s prowess in coding, will it completely replace programmers? Or how much longer will it take? Therefore, I’m curious about the most complex web project everyone has ever worked on, and whether AI can complete it independently at present.
1 Like
When AI completely replaces programmers, you have created Skynet, and the terminators are coming.
2 Likes
Do you mean that programmers have killed the profession of programmers?
No, my point is that AI cannot completely replace programmers. Ever. If it does, you have created self-generative AI, which is when the AI decides it doesnt need humans anymore and the terminators come to fix the “problem” that is humanity.
One of the most complex web projects I’ve worked on was building a niche-based content and gaming platform. It involved structuring multiple pages, optimizing for SEO, managing content, and making sure the user experience stayed simple even with different sections.
The challenging part wasn’t just development, but handling everything together — like keyword research, on-page SEO, content creation, and basic UI improvements. Also, making the site load fast and mobile-friendly took a lot of effort.
Coming to AI — yes, AI can definitely help in building such projects, but it can’t fully replace human effort yet. AI tools can:
Generate content
Suggest designs
Help with SEO and keywords
Automate repetitive tasks
But the real challenge is strategy, creativity, and decision-making, which still needs human thinking. AI doesn’t fully understand your audience, niche, or long-term goals the way a person does.
So in my opinion, AI is more like a powerful assistant. It can speed up the process a lot, but building a successful web project still requires human input to make it actually work and grow.
Limiting the question to web applications ignores the vastness of back-end systems. Web applications should not do processing that can be done later.
Common examples of complex applications are operating systems, compilers and AI. Embedded systems such as flight control (in the air and on the ground) can be complicated. Self-driving cars require complexity. Simulation software such as for weather prediction is complicated.
AI was used to write a C compiler, totally autonomously. It is not the best compiler but it is quite functional.
The greatest limitation is the ability (by people currently) to describe what needs to be done.
Problem is, that explaining the AI detailed enough what you need to get useful output, takes longer then doing it by yourself….
If you have the knowledge to do it by yourself you are faster doing it by yourself, then explaining the AI and after control what the AI has done and fix the problems. Have you ever checked the code AI produces? It’s a heavy mix of very old and a little bit new code. Most off it completed outdated. I still get most of the JavaScript code with var declarations instead of let and const. Did you ever run a SAST scan over an AI produced code? After AI has produced the code you need another AI to check the code against vulnerabilities.
Yes the code is as good as it is when you hire a cheap Asian company to write it, but it is far from what is state of the art today.
I really don’t understand the hype of all this AI written applications. Yes they work, but if you just focus on functionality and nothing like security and maintainability, you might pay a high price …
By the way: how will you be sure the AI is not creating a standard backdoor in your code for the NSA or other similar institutions? What do you think how long it will take since the hackers in the world find a way to get into all apps created by the same AI when this AI has a security problem?
An interesting idea, that conventional, manual programming is exactly that. You are instructing a computer system on what you want it to do in a very detailed and unabugious way.
So to instruct an AI to the same level of detail, with unambiguous commands that won’t be misinterpreted, could be just as much work as programming it yourself.
That is assuming you have already taken the time to learn the programming language, which I’m quite sure is what all the AI users are doing. 
1 Like
I think that systems design taught by most colleges teaching software development teach that avoiding specifications can be highly costly. I think they teach that any mistake made in design or specifications is likely to cause a significant amount of time to fix. In a large project with multiple developers, software is supposed to be documented. The specifications can be useful documentation.
About a month ago in these forums I posted specifications for a simple static site generator. Originally the specifications seemed simple. I asked for reviews by AI and they provided extensive suggestions and corrections much like a person would. The specifications were highly improved in the process. When I was satisfied with the specifications I got AI to create the program, it did so nearly instantly. In that instance, I developed specifications much like developing source code and AI generated the source code much like a compiler generating machine code. I probably could have written the program faster if I had just done it but it would not have had the improvements until after I spent much time.
In the past I had a project wherein I wrote a COBOL program to parse out the tools and materials in manufacturing instructions (written for use only by people) for black-world aircraft. The manufacturing engineer was not a programmer but he wrote very useful specifications. The project was much easier to complete due to his clear instructions.
AI will say:
most modern AI models default to ES6+ standards (
letandconst) because they are trained on modern, high-quality codebases.
AI will tell you that they can make mistakes. You are correct, we should not trust that the generated code does what it is supposed to do without mistakes. We need to study the code.
1 Like
For the record (and to get back to the original point of the thread kinda…), this is why I said AI cannot completely replace programmers.
The simple truth is that “vibe coders” think they can stick their specifications, no matter how well worded, into an AI, take what comes out, and it will be perfect. And even if they doubt that it is perfect, they lack the ability to verify this for themselves, or correct it if it requires change. They can’t even explain how the code does what it does. So a programmer will be required, even if it’s just to follow up behind the AI and fix things. Or even to EXPLAIN things.
2 Likes
Yes the original post says completely replace programmers and I responded to the suggestion that it is useless to use AI for any coding.
Built an AI resume builder that basically rewrites your resume to match job descriptions—pulls out keywords, reorganizes your experience, the whole thing. Hardest part was getting it to actually understand what you’re saying instead of just doing a find-and-replace with fancy words, plus dealing with all the weird formatting people use. Could AI handle it solo? Maybe, but honestly you’d still need someone to actually build the thing and make sure it doesn’t break constantly lol.
1 Like
I believe it’s only a matter of time before AI replaces all programmers. The question is, how can programmers in the current window period ensure their transition to product managers?
1 Like
The most replicated project I have developed is an online video editing tool similar to CapCut. I believe that, at least for now, AI cannot be replaced, but the future is uncertain
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.