For a new project I like to start by writing code that sketches out ideas of how the software should hang together.
TDD at this point is kind of wasteful because if you decide to backtrack or rework aspects of the design then you have to also rework all of the tests. I find that once the design begins to crystalize, adding unit tests at this point is valuable, but not necessarily in a 'test-driven' way.
TDD by itself, per se will not automatically produce a good design, and can often produce bad designs with huge proliferation of classes / interfaces that makes a codebase more complex and difficult to understand.
TDD is not the be all and end all that some commentators seem to believe.
TDD at this point is kind of wasteful because if you decide to backtrack or rework aspects of the design then you have to also rework all of the tests. I find that once the design begins to crystalize, adding unit tests at this point is valuable, but not necessarily in a 'test-driven' way.
TDD by itself, per se will not automatically produce a good design, and can often produce bad designs with huge proliferation of classes / interfaces that makes a codebase more complex and difficult to understand.
TDD is not the be all and end all that some commentators seem to believe.