To develop an algorithm for any application takes thought and a lot of trial and error. The algorithm must be coded, compiled, tested for compliance with the specification. If it does not perform to target, the code must be amended, recompiled and tested again. The process is cyclic and time consuming. In this paper a novel method is introduced which allows the building or tuning of algorithms or programs at run-time by using an active database. The paper uses the domain of robotic vision as a case study to introduce the concept, particularly the first stage of the object recognition process known as segmentation i.e. extracting the primitive characteristics of the objects of interest. The system has been implemented upon the REFLEX active database system.