The video examines the limited effectiveness of OpenAI’s new GPT-5 prompt optimizer tool, showing that while it can improve clarity and structure in some text-based prompts, it often fails to significantly enhance complex outputs like coding and lacks interactivity. Ultimately, the creator argues that prompt engineering remains relevant but requires thoughtful design beyond relying on automated optimization tools, urging viewers to stay critical of AI hype.
The video explores the current relevance of prompt engineering in the age of advanced AI models, specifically focusing on a new prompt optimizer tool introduced with GPT-5. The creator tests this official OpenAI tool by comparing outputs generated from unoptimized prompts versus those refined by the optimizer. Interestingly, the results sometimes vary significantly, with the AI providing completely different answers to the same question depending on the prompt’s optimization. This raises questions about the consistency and reliability of prompt engineering as a practice.
GPT-5’s release is described as underwhelming, especially given the hype around it. Alongside the model, OpenAI provided a prompting guide intended to help users craft better prompts, but few people actually read it. To address this, OpenAI introduced a prompt optimizer tool that automatically refines user prompts. However, the creator finds the tool somewhat lacking, noting that it does not interact or ask clarifying questions and simply reformats the prompt in a somewhat basic way. The tool’s interface could also be more user-friendly, such as having a direct “optimize” button within the chat interface.
The creator tests the optimizer on different types of prompts, including one for generating a game and another for summarizing the history of digital IDs in the UK. While the optimizer adds structure and clarity to the prompts, the improvements are modest. In the coding example, the AI-generated code contains bugs and errors, indicating that prompt optimization alone does not guarantee high-quality outputs. The creator expresses frustration with the AI’s declining performance and the optimizer’s inability to significantly enhance results in complex tasks like coding.
In another example involving tech journalism content, the optimizer successfully condenses and clarifies a lengthy prompt, making it more manageable for the AI to process. This leads to a more coherent and relevant response compared to the original prompt. However, the creator notes that the optimizer’s summarization might omit some specific details that could be important for pattern recognition. Despite this, the optimized prompt yields a better answer, demonstrating that prompt optimization can be beneficial for text-based queries, even if it is not perfect.
Overall, the video concludes that while prompt engineering is not dead, the new GPT-5 prompt optimizer tool offers a mixed bag of results. It can improve clarity and structure in some cases, especially for text prompts, but it is not a magic solution and does not replace the need for thoughtful prompt design. The creator suggests that the tool should be more interactive and integrated for better usability. They encourage viewers to stay updated on AI developments and remain critical of the hype surrounding new releases.