Authoring hypertext documents has always been a difficult task. The cognitive overhead problem, well known from authors when creating more or less structured hyperdocuments, has been tackled in a variety of forms. In this paper we propose that capture-based systems should support flexible hypertext structures generated by linking by interacting operations. These operations include operations that capture user-interactivity during a live session - which we call linking by capturing operations - and allow the captured contents to be extended with complementary operations available outside the live session - which we call linking by augmenting operations. We demonstrate by example how such a combination leads to the generation of flexible hypertext structures by presenting our implementation in the educational domain.