Meta-Prompting and Agents: Getting the Most Out of ChatGPT

In the talk, we will discuss the principles behind Large Language Models (LLMs), how models accept responses, and the possibilities for improving response accuracy. We will dive into the latest and non-obvious techniques for interacting with ChatGPT (such as Meta Prompting and the use of agents), increasing response quality as if we already had access to ChatGPT-5. Through practical examples, we’ll examine how techniques like Chain-of-Thoughts and One-Few Shots can significantly impact response quality depending on the prompt. We’ll also discuss how to leverage LLMs in different roles to tackle tasks from multiple angles, engaging various virtual experts. Real-world examples will demonstrate how these methods can assist in requirements analysis and test model preparation.