TDD encourages better APIs and more testable code. I don't use TDD to prevent more bugs. I don't even know if this is really true.
Everytime I think of a new feature I fastly have an API in mind. Starting with a test almost always shows me a better way to provide the functionality (easier, more failsafe). Thats one of my two key features of TDD.
The second key feature is that the tests you write are much more expressive than tests you write afterwards. I observe that everyday in my team and company. Writing tests after the implementation is done leads to a lot of mocks and just a static verification that the implementation works how it works. In the end, it is difficult to change the implementation and easy to break the specification. Writing your test ahead of the implementation means you test the specification and not how you implemented it. Tests become much easier to read as they express the specification instead of a 1000 lines test where every variable assignment of the implementation is checked.
As a last hint: I read here that people complain that TDD does not work with "prototyping". I agree with that - the problem with that statement is that a lot of people use their prototype as productive code later on instead of rebuilding it after the experimental phase. In my opinion that is a misuse of the word "prototype". A prototype is an experimental work to verify or learn some ideas and afterwards designing a better system.
Other people may state that testing everything is not possible - that there are situations where it is not possible to write a test. I would bet that in 90% of that cases, this is just wrong and means a lack of testing experience of the writer. I can really recommend the following book: http://www.amazon.com/Driven-Development-Embedded-Pragmatic-... I rarly do embedded system programming. My main focus are languages like Java and Go but this book is worth its money independent of your choice of programming language.
Fresh book recommendations delivered straight to your inbox every Thursday.