AI writes your API in five minutes. What do you bring?
In 2026 you open Claude or Copilot, type "build me a Laravel REST API with authentication" and five minutes later you have something running. Great. Now one question: did you actually read that code, or did you just check that it ran?
You didn't write it and you didn't decide any of it. If it falls over in production tomorrow, your only move is reopening the same chat and begging the model to sort it out, because you know nothing about that code that you didn't already know before generating it.
The question that matters is what you put in it. If the answer is "the prompt", your contribution is a request anyone in the world can write for a fifth of your salary. At that point the story about AI replacing us sooner or later no longer concerns you: you're already replaced. Your company just hasn't put it in writing yet.
And if you're a junior or mid dev, the problem is doubled. A senior spots a query that won't hold at a glance, smells a leaky validation from a mile away. You don't have that eye yet, and if you delegate the thinking to the model too, you never will. You're gambling away exactly the years you should be building it in.
The alternative is not giving up the model, it's refusing to take its output on faith. It generates, you decide whether that query survives real traffic, whether that validation covers the input nobody thought of, whether that authorization rule lets one user read another user's data. Those are questions the model won't ask on its own. You have to ask them, and you can only ask them if you know they exist.
Writing code is worth nothing anymore, because anyone can do it in five minutes. What's worth something is the thinking you put into it. If you put in none, the most expensive and slowest link in the production chain is you.
That's why I wrote a book: to put you back on the deciding side. The model writes a single endpoint very well, but an API is a chain of decisions that condition each other, and asked for one piece at a time that coherence is exactly what you lose. So the book follows one API from start to finish, from a clean Laravel install to a deploy on a VPS. The domain is deliberately boring (a catalog of books!) so you spend your time understanding how an API is put together, not decoding the business. Inside are the things you need to be able to judge when the model generates: what your status codes are actually telling the client, where validation ends and authorization begins, what breaks under load once the data gets real, what tells you the work is actually done. The last chapter is about Claude Code, and it sits at the end for a precise reason: after nineteen chapters the model works for you. Read only that one and you stay someone who works for the model.
If you've been working with Laravel for years and you move through Policies, queues and Resources with your eyes closed, this book isn't for you. If instead you're looking for a way to churn out APIs without having to understand them, that book exists, but I didn't write it.
The chapter on eager loading and N+1 is online for free, in full:
antonio.popolizio.it/laravel-rest-apis/sample.pdf
That's the point where most Laravel APIs fall over once the data gets real. No form, no email: you download it and that's it. The rest is on Amazon: antonio.popolizio.it/laravel-rest-apis.
I'm curious how many people here actually read generated code line by line, and how many just check that it runs.