Since the release of LLM-based tools such as GitHub Copilot and ChatGPT the media and popular scientific literature, but also journals such as the Communications of the ACM, have been flooded with opinions how these tools will change programming. The opinions range from ``machines will program themselves'', to ``AI does not help programmers''. Of course, these statements are meant to to stir up a discussion, and should be taken with a grain of salt, but we argue that such unfounded statements are potentially harmful. Instead, we propose to investigate which skills are required to develop software using LLM-based tools. In this paper we report on an experiment in which we explore if Computational Thinking (CT) skills predict the ability to develop software using LLM-based tools. Our results show that the ability to develop software using LLM-based tools can indeed be predicted by the score on a CT assessment. There are many limitations to our experiment, and this paper is also a call to discuss how to approach, preferably experimentally, the question of which skills are required to develop software using LLM-based tools. We propose to rephrase this question to include by what kind of people/programmers, to develop what kind of software using what kind of LLM-based tools.