EN

95% of People STILL Prompt ChatGPT-5 Wrong

#AI Tools
2025-11-25 18:399 min read

Content Introduction
Ask Questions

The video discusses the impact of ChachiPT5's launch, particularly how users are reporting worse results despite the model being more powerful. It explains a fundamental change in GPT-5's architecture that affects old prompting techniques. The speaker shares five tips to optimize prompts for better results, beginning with increasing detail to direct the model effectively. It emphasizes the importance of verbosity control, structuring prompts with XML tags, and utilizing OpenAI's official prompt optimizer tool. The video further elaborates on creating a consistent quality improvement loop with tips to refine user interactions with GPT-5. It concludes by encouraging viewers to explore deeper strategies for effective prompt engineering.

Key Information

  • After ChachiPT5 launched, millions of users reported receiving worse results despite using the same prompts as before.
  • ChachiPT5 is a more powerful model, so it was expected that the same prompts would yield better results, but OpenAI made significant changes to GPT5's architecture.
  • The algorithm has an invisible router that selects the model handling user requests, which sometimes leads to suboptimal results due to its inefficiency.
  • Users can improve their interactions with GPT5 by incorporating five specific tips to enhance output quality.
  • Two major updates to GPT5 include model consolidation, reducing the number of models from several to just three, and improvements in instruction-following capabilities.
  • Users should reformulate their prompts as the previous techniques may perform worse with the new model.
  • Encouragement to create more structured prompts, use verbosity controls, and apply the perfection loop method for iterative improvement of outputs were discussed.
  • OpenAI's prompt optimizer tool can enhance prompts for GPT5, yielding better outputs for various tasks.
  • Examples were provided on optimizing prompt length and organization using XML tags to ensure clarity and structure.

Timeline Analysis

Content Keywords

ChachiPT5

After the launch of ChachiPT5, many users reported receiving worse results despite unchanged prompting methods. This is attributed to fundamental architectural changes in the GPT5 model.

GPT5 architecture

The architecture change in GPT5 requires users to adapt their prompting techniques, as the previous methods perform worse with the new model.

OpenAI guides

The author tested GPT5 for a month using OpenAI's official guides and identified five simple tips that significantly enhance output quality.

Router nudge phrases

By adding specific phrases at the end of prompts, users can effectively nudge the invisible router to select a higher reasoning model.

Verbosity control

Users can control the verbosity of GPT5's output through specific phrases to achieve desired lengths and details in their responses.

XML sandwich

Structuring prompts using XML tags can help in organizing instructions better for GPT5, improving comprehension and output quality.

Perfection loop

The perfection loop advises users to allow GPT5 to create its own definition of excellence and self-grade, leading to improved results.

OpenAI prompt optimizer

OpenAI's prompt optimizer tool rewrites prompts for better performance and helps users learn to create effective prompts themselves.

Prompting techniques

Incorporating the learned techniques, such as XML structuring and the perfection loop, ensures high-quality outcomes from GPT5.

Related questions&answers

What happened after ChachiPT5 launched?

Millions of users reported getting worse results even though they haven't changed how they prompt.

Why are the results worse with ChachiPT5?

OpenAI made a fundamental change to GPT5's architecture, which means old prompting techniques now perform worse in the new model.

What is the problem with the new model?

Despite being a more powerful model, 95% of users don't know that the same prompts should work better but sometimes yield worse outcomes.

What are some tips to improve output with GPT5?

I discovered five simple tips that drastically improve outputs for users.

What is model consolidation in GPT5?

Previously, users had access to different models; now, there are just three: GPT5, GPT5 thinking mini, and GPT5 thinking.

What changes were made to GPT5's ability to follow instructions?

GPT5 is much better at following instructions, but this can be a double-edged sword as it's worse at guessing expectations when prompts are vague.

What should users do to optimize their prompts for GPT5?

Users should incorporate router nudge phrases in their prompts to guide the model toward better reasoning.

How does verbosity control impact outputs?

Verbosity control settings can help the model tailor the length of its responses based on user instructions.

What is the XML sandwich for structuring prompts?

The XML sandwich involves using tags to organize prompts, helping the model understand distinct sections of information.

How can the perfection loop help with complex tasks?

The perfection loop encourages continuous improvement, where the model critiques and refines its own outputs until they meet a defined standard.

More video recommendations

Share to: