Matteo Cortese

← Back to the shelf

Beyond Vibe Coding

Addy Osmani

The argument that settled a question I kept turning over: AI doesn't make engineering judgment less important — it makes it the only thing that actually matters.

“Vibe coding” is fun right up until it ships to production. Osmani’s argument is that the AI era doesn’t make engineering judgment less important — it makes it the whole job. Anyone can prompt a model into some working code. Knowing whether that code is correct, safe, and maintainable is where the value actually moves.

The book is organised around a simple but important observation: the bottleneck in software development has shifted. For years it was implementation — how fast you could write working code. AI has largely dissolved that constraint. The new bottleneck is judgment. Understanding the problem clearly enough to evaluate whether the generated solution is actually right. Knowing when to trust the output and when to push back on it. Designing systems that won’t fall apart when the prototype becomes a product.

What I found most useful is the way Osmani treats the skills that remain irreplaceable. Architecture, code review, security thinking, testability, observability — none of these get easier because the model writes fast. They get harder to skip because the speed of generation outpaces the speed of understanding. You can have a working codebase in an afternoon and no idea what it actually does. That gap is dangerous in a way that slow hand-written code never was.

I read this as an AI product builder, not a full-time software engineer, and it still landed. A lot of my work involves evaluating what AI tools produce — choosing what to ship, what to trust, what to refactor before it becomes a liability. The frame Osmani builds is directly applicable to that role. Coding becomes cheaper; deciding what code to build, how to structure it, and what failure looks like becomes more expensive.

Pair this with Claude Code for Product Managers on the same shelf. One looks at the shift from the product side, this one from the engineer’s. Both point at the same place: the people who will do well in the next few years are the ones who can build with these tools and still understand what they’ve built.

If you’re a developer wondering whether AI makes your skills obsolete, this is the most direct answer I’ve read. It doesn’t. It just moves the valuable ones higher up the stack.