You can easily write the test after the code, and then use git interactive rebase to reorder them, during what you can see that the test fails without the code change. "First" means nothing. Any way of demonstrating that a test fails without the code change and passes with the code is valid regardless of order of development.

Say you're working on a programming language. You add a new feature, and then a passing test for it. (Wrong order according to TDD).

But you can easily check that the test fails using an installation of the current release version of your language, which doesn't have the change.

I.e. test passes with ./mylang; fails with mylang (found via PATH in /use/local/bin).