Over the years, developing tools and products has changed a lot in the web development world. We have changed our tools, languages, concepts and philosophies over time and I though it would be interesting to lean over this topic. We are going to go through a pretty large timeline, from the early beginning of web (2000s), going through the frameworks war and a bunch of other things, until what we know today with AI.
The MVC Revolution : Organizing the chaos
Back in the old days, web code was "spaghetti". You’d have PHP tags mixed with HTML, which was mixed with inline CSS, which contained a SQL query. It was pretty tough to maintain and we can safely say that the MVC model was a really helpful thing at that time. It would abstract things to organize the code into 3 blocks :
- Model -> Abstraction of the database. You would stop writing SQL queries by hand, and started calling things like
Users.all(). - View -> Abstraction of the UI. Templates meant you didn't have to
echoHTML strings. - Controller -> Abstraction of the logic. The bridge that kept the data and the UI from touching each other.
It's in this era that is actually born the term "Web developer". But it was not that popular yet. Design patterns started to be a thing and we began to optimize the way of building web app instead of just "making pages load".
The SPA Revolution : The browser as an OS
Around 2011, we realized that re-rendering the entire page when a user clicked a button was slow. That's when we invented the concept of Virtual DOM, client-side rendering and Single Page Application. It's also at that time that frameworks like Angular and React were born. These technologies actually introduced a massive layer between the developer and the actual browser DOM.
Developers stopped being "Document Creators" and became "Application Developers". The complexity shifted from the server to the browser and suddenly, you needed to build your application as a bundle with webpack just to run a Hello World.
We didn't know back then that the Javascript ecosystem would turn into such a quagmire.
The container era : Death of "It works on my machine"
As applications and services grew up, the problem became the deployment. We all know the phrase : "It works on my machine". That's when things like Docker came in. If every devs in a team had a different version of node, python or a linux library, you could create a Docker container so that everybody was on the same page.
"Devops" became a thing even though nowadays, every developer should know a substantial amount of devops to work in a team. We stopped installing software on servers. Instead, we started "shipping the whole computer." And the abstraction is now the hardware ; it became invisible. You were just moving boxes around a global grid.
The Serverless era : The vanishing backend
But recently, even "managing a container" started feeling like too much work. Why should a developer care about CPU cores or memory allocation? Now with things like AWS, Azure and Google Cloud platform, you hand over the entire execution lifecycle to the cloud provider. Your whole infrastructure now depends on an external actor.
We came to a point where we abstract entire logic blocks of our application. You put the state of your app into external databases and S3 buckets. We also now have a problem of "Black Box" debugging : When your code is slow, is it your code? The database? Or is it a "Cold Start" in the provider's infrastructure? You can't ssh into the server to check because the server doesn't "exist."
The AI layer : Abstracting the code
And of course, the final boss : AI, or rather LLMs. Or agentic coding. Or prompt engineering. I don't know, I feel like there is a new thing every tuesday. Now we abstracted one of the most important parts of being a developer. Writing code (I didn't say good code).
Thanks to hype bros coming from the NFT trends, you can believe that you're a software engineer by just prompting Claude code to build a to do list application. I guess that the bottleneck was the human's ability to write syntax. How could we build good products without AI...
The cost of abstraction
I think that this is interesting to take a step back to see and remember where we came from, and to see where we are now. All these layers of abstraction have been created for a reason (for the most part) : helping the human in his task. But the problem is that the more you abstract things, the more control you lose. It then becomes really hard to truly understand what's going on. The Leaky Abstraction principle taught us that when things break today, they break in ways that are incredibly hard to debug on large scales. Our day-to-day work, now looks like an agglomerate of concepts, dependencies and libraries written by other people, with so many abstractions. So much so that we tend to forget how things work under the hood. Especially for the young and junior programmers. They have the keys to the kingdom, but they may have never seen the foundation it’s built on. I know this because I started my programming journey during the frameworks war era. And I think that I didn't take the best learning path. Like why do I need to learn React and Vuejs when I barely even know Javascript ? And what the hell is Docker and containerization ?
All this to say, learn the basics and take your time folks. AI is not gonna take your job, even if some people would want this to happen.