The context of this research is initially, an order from a private school in Geneva, to give a few hours of continuous training to teachers. It is a primary school, with children from 4 to 12 years old. The request concerned the organization of the teaching of problems, called "open problems". The project started in September 2017 and is still running until June 2019 or more. It is a longitudinal study. Two years to implement the teaching of problem-solving. My part of researcher is to understand how teachers and students cope with these problems. A first question to be problematized is "If we practice open problems from early childhood, can we measure an impact in grade six, or even before, on the students' reasoning". The first results I will present concern the following difficulties, for example: • implement what we call the devolution of the task, which appears as a long process to appropriate the task, • manage the organization of the sheet space for the pupils, • etc. Among these difficulties, I will highlight the positive contributions such as the fact that the students had fun, participated well and learned about maths. For the teachers, the most important and, at the same time, the most difficult, is not to induce the approach, leaving students to experience the tasks by themselves. We will see that letting ignorance "float" can be an interesting way to support learning. The aim of the study is to understand both how to promote reflection and to improve learnings.