This is how Facebook was introduced in 2004 — or sometime around then.
Look at the design. Look at the UX decisions. Or better yet, look at the technical choices. The front end was built with HTML, CSS, and JavaScript; the back end ran on PHP; the database was MySQL; everything was served on an Apache server — and all of it was built in just one or two weeks of work.
Today it looks primitive. But at the time, it was modern. It was cutting edge. It was a product that changed the world.
Around the same years, Amazon’s user interface was significantly redesigned, simplifying navigation down to a few essential choices, and it looked roughly like this:
Nothing impressive by today’s standards, but at the time it was a massive system with complex infrastructure, supported by thousands of servers and millions of users.
And it was a revolution in the world of email. The storage offered: 1000MB. At the time, Hotmail or Yahoo offered 2–6MB. The difference was not in percentages, but in multiples. JavaScript + AJAX solutions made it possible not to refresh the page every time you opened an email, and it felt like the future, like something that would rewrite the rules.
Also around 2004, eBay was already an infrastructure giant with a product like this:
Millions of auctions, real-time bidding, a global audience, PayPal as the most advanced payment solution — and it was the highest-quality technology available on the market.
All of these products were modern and technologically cutting-edge at the time. I used them myself, and I didn’t have even the slightest doubt that this was the pinnacle of quality, engineering excellence, and ingenuity.
Now let’s imagine a fiction.
At the end of 2005, ChatGPT appears. It is trained on all these cutting-edge systems — Facebook, Amazon, Gmail, eBay. It sees PHP, MySQL, Apache, Oracle, C++, AJAX. It sees the best architectural solutions that existed at the time. It sees what is the technological peak of the world.
And the same thing begins that is happening today.
Everyone explains that programmers will no longer be needed. That creating a new eBay is a matter of minutes. That it’s enough to write a prompt and the system will generate the entire application. Everyone builds applications, everyone generates code, everyone combines what already exists. Everything looks advanced. Everything looks like it’s changing the rules of the game.
But what does that 2005 ChatGPT actually create?
It optimizes PHP. It generates better MySQL query templates. It makes nicer HTML. It combines AJAX with what already exists. It creates combinations from what it sees around it. It does not cross the boundaries of its era, because those boundaries do not yet exist in its training data.
And then, somewhere in parallel, people begin working on Go and present it publicly in 2009. A language that changes thinking and simplicity in infrastructure code. In 2009, Node.js and npm appear, changing the server-side JavaScript model. In 2009, MongoDB appears, questioning the dominance of relational databases. In 2013, Kubernetes appears, fundamentally changing how we manage containers and distributed systems. In 2015 — React Native, and so on, and so forth.
Now let’s think critically.
In 2005, with ChatGPT available, would Kubernetes have appeared? Would the Go language have appeared? Would a document-oriented NoSQL database have appeared if the model had been trained only on the dominance of relational systems? Would a new asynchronous server model have appeared if the standard at the time had been completely different?
The answer is very simple. AI would not have created it. A human would.
Because it does not yet exist in its training data. It cannot generate what the world has not yet invented. It can combine, optimize, accelerate — but it cannot create an architectural breakthrough out of nothing if there are no precedents for that breakthrough.
Yes, it is a fact that with AI some technologies would have appeared faster. But they would not have appeared because AI invented them. They would have appeared because humans questioned what at the time seemed self-evident.
And now let’s look at today. Everything that today seems modern, advanced, intelligent, and inevitable will, in 20 years, look just as simple and even slightly amusing as the 2004 eBay homepage looks today — even though it was the flagship of technology. And perhaps in 20 years someone will say: “Can you imagine, they seriously believed that prompts could replace all of engineering.”
Technologies change. Tools change. AI will also change. But architectural breakthroughs are still created by people who begin to think differently than the data of their time allows. And here the question is not whether AI will replace programmers. The question is who will create the next Kubernetes.
chatGPT of 2005 - https://itknyga.lt/2005-chatgpt/