The article discusses the emerging trend of Large Language Models (LLMs) revealing their reasoning processes. 

DeepSeek, a new LLM, exemplifies this by openly displaying its chain of thought, a feature some experts see as beneficial for user prompt refinement and bias identification, while others view it as a mere marketing tactic.

The author argues that transparency fosters iterative improvements in prompt engineering, leading to better results. Conversely, some believe concealing the reasoning process promotes more independent "thinking." The debate centers on the balance between transparency and the potential for anthropomorphizing AI.

This article is also available in English on my Medium page, «When it comes to using an LLM, knowing how its thought processes work is a big advantage«