In this talk, the speaker discusses the concept of self-conscious reflexive interpreters, inspired by John McCarthy's 1959 paper on an advice taker program. The goal of an advice taker is to build a problem-solving software that can learn from its experience, accept advice from an external entity, and communicate with that entity. The speaker explores the features proposed by McCarthy, such as representability of all behaviors in the system itself and extensibility, as well as the challenges and limitations faced in building such a system. The speaker also mentions the work of other researchers like John Doyle and Doug Lenitt, who were interested in similar areas. The speaker introduces their own project, the Barlament Advice Taker or BAT, which is a system that uses the mini-kanran constraint logic language to solve various programming problems. They discuss the idea of using BAT as an external solver and the integration of heuristics and meta-heuristics in the system. The talk concludes by acknowledging the early stage of the project and highlighting the importance of exploring different approaches to building an advice taker system.