This is a genetic algorithm (GA) based system which is under development at The University of Birmingham, UK. The version used discovers conjunctive normal form (CNF) or fuzzy CNF classification rules for pre-classified data. The system uses a number of metrics to evaluate the worth of rules. Genetic algorithm operators are then used to generate a new population of rules from the best rules in the previous population. This process of evaluation and rule generation is repeated until a termination criteria is met. Whilst some aspects of the work are novel, it is not currently a particularly powerful KDD system. Future work will include new operators and metrics which make use of domain knowledge and the addition of other forms of knowledge discovery, such as the generation of association rules. However, the advantage of the Haiku system is that it can visualise any KDD or machine learning approach.