Angular is known for providing developers with a high level of abstraction. Router, HttpClient, Signals, Forms, and many other tools allow us to build complex applications without constantly thinking about browser internals or the JavaScript runtime.

That convenience runs into an important software engineering principle Joel Spolsky described years ago: The Law of Leaky Abstractions.

The law states:

All non-trivial abstractions, to some degree, are leaky.

In other words, an abstraction can hide complexity, but it cannot eliminate it.

Example 1: Angular Signals

Signals offer a simple, intuitive way to manage application state. At first glance, it feels like you no longer need to think about how state changes propagate or when rendering happens.
Then a component starts re-rendering more often than it should, or a computed() recalculates in a place you didn't expect. Suddenly you're reading about Angular's dependency tracking and change detection scheduling to figure out why.

Example 2: HttpClient

Angular's HttpClient makes sending HTTP requests incredibly straightforward.

However, when we encounter CORS errors, request timeouts, authentication issues, or caching problems, we quickly find ourselves diving into the details of HTTP, browsers, and network communication.

The abstraction has leaked.

Example 3: RxJS

RxJS makes asynchronous data streams feel manageable. A handful of operators and you can express HTTP calls, user interactions, and data transformations declaratively.
Until a race condition shows up. Or a request gets canceled when you didn't expect it to. Or switchMap does something different from what mergeMap would have done, and you can't explain why. At that point, the only way forward is understanding the event loop and how JavaScript actually schedules async work. RxJS won't save you from needing that.

AI as a New Leaky Abstraction

In recent years, we have gained another powerful layer of abstraction: AI-powered development tools.

Today, we can generate Angular components, write RxJS pipelines, create tests, or even implement entire features with the help of AI assistants. These tools significantly improve productivity and accelerate development.

However, the Law of Leaky Abstractions applies here as well.

When AI-generated code does not behave as expected, when performance issues emerge, or when maintainability and security concerns arise, we must still understand what is happening beneath the surface.

AI does not eliminate the need to understand Angular, TypeScript, browsers, or software architecture. In fact, the more we rely on AI, the more important it becomes to understand the fundamentals so that we can evaluate the quality of the solutions being generated.

AI may be the most relevant modern example of the Law of Leaky Abstractions. It allows us to operate at a higher level than ever before, but when things go wrong, we still need to step down a level and understand the underlying system.

What Does This Mean for Developers?

Using a framework does not free us from understanding the fundamentals.

Today, the same is true for AI. Using AI tools does not remove the need to understand the technologies we work with every day.

Senior engineers often stand out because they understand what lies beneath the abstraction and can identify the root cause when things stop working as expected.

Angular makes us productive. AI makes us faster. Neither one makes JavaScript, the DOM, networking, or basic software engineering optional. They just make it easier to forget that, until something breaks.