The primary purpose of this paper is to stop people using the Task Analysis for Knowledge Descriptions (TAKD) method. Secondly, by describing the history of TAKD's development and demise over nearly two decades, it allows lessons to be learned that may be relevant to existing methods and to those being developed. Both the adequacy of TAKD as a product and its delivery to HCI and software engineering practitioners is examined, and TAKD's failure on both these aspects is described. The value of developing software tools to support analysts is emphasised and illustrated by the development of the LUTAKD toolkit. It is argued that it is essential for task analysis to be able to model software if it is to be used as part of a software requirements and design specification process.