The arrival of ChatGPT has caused a lot of turbulence also in the field of software engineering in the past few months. Little is empirically known about the capabilities of ChatGPT to actually implement a complete system rather than a few code snippets. This paper reports the first-hand experiences from a graduate level student project where a real-life software platform for financial sector was implemented from the scratch by using ChatGPT for all possible software engineering tasks. The main conclusions drawn are as follows: 1) these findings demonstrate the potential for ChatGPT to be integrated into the software engineering workflow, 2) it can be used for creating a base for new components and for dividing coding tasks into smaller pieces, and 3) noticeable enhancements in ChatGPT-4, compared to ChatGPT-3.5, indicate superior working memory and the ability to continue incomplete responses, thereby leading to more coherent and less repetitive dialogues.