How to Prompt GPT-5

Key takeaways

Explicit and structured prompts yield better results with GPT-5, requiring a shift in how users interact with the model.
Utilizing foundational techniques such as planning phases and asking the model to 'think harder' encourages deeper outputs and better engagement.
Avoid conflicting instructions in prompts to prevent GPT-5 from spending excessive reasoning tokens trying to resolve contradictions.
Leverage the model's iterative capacity and agentic toggles to enhance output quality and efficiency in task handling.
Metaprompting can be an effective strategy to improve prompt clarity and efficacy by soliciting the model's suggestions for better prompts.

Summary

Introduction to GPT-5 Prompting Techniques

The episode kicks off with a brief overview of the GPT-5 model which has created a stir amongst users regarding its performance. The host notes that, almost a week into the GPT-5 era, the community remains divided over its efficacy, claiming it to be more steerable but also recognizing that users have to adapt their prompting strategies in order to optimize results. The episode provides 11 prompting techniques that stem from various expert insights to navigate this new model.

Foundational Techniques for Prompting

In this section, the focus is on foundational prompting techniques essential for leveraging the power of GPT-5. The host emphasizes that GPT-5 requires clarity in prompts, encapsulated in techniques such as asking it to 'think harder' and introducing explicit planning phases. Recommendations also include structure in prompts to delineate expectations clearly, as well as minutiae processes that guide the model through tasks effectively.

Advanced Prompting Strategies

The episode delves deeper into advanced prompting strategies, such as metaprompting—where users can request the model to suggest improvements to the prompts given. This is underscored by insights from OpenAI telling users to avoid conflicting instructions, as it can hinder the model’s performance significantly. Another innovative approach discussed includes leveraging the model's capacity for iteration to enhance output quality.

Leverage Agentic Capabilities

The host explains the agentic capabilities embedded within GPT-5, which should be utilized for tasks that can be done concurrently. Being explicit about the parameters for reasoning effort and verbosity can enhance the outcomes of interactions with the model. The importance of effectively using toggles in guiding the model towards parallel processing and controlling output length is also discussed.

Conclusion and Takeaways

The episode wraps up summarizing the main points discussed and encouraging users to adopt these new prompting techniques to optimize their experience with GPT-5. The host invites listeners to share their experiences as they try out these techniques, further contributing to the evolving understanding around GPT-5.

Show notes

Nearly a week into the GPT-5 era, users are still divided on its quality—but one thing’s clear: it’s more steerable than any previous model, and prompts make or break results. In this episode, we cut through the debate and share 11 practical prompting techniques you can use right now to get more from GPT-5. From “think harder” prompts and explicit planning phases to structured formatting, avoiding conflicting instructions, leveraging iteration, and meta-prompting, these tips draw on insights from OpenAI, early testers, and top prompt engineers. If you’ve been struggling to get GPT-5 to deliver, this guide will help you turn frustration into high-quality outputs.

Brought to you by:

Plumb - The automation platform for AI experts and consultants ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://useplumb.com/⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614Subscribe to the newsletter: https://aidailybrief.beehiiv.com/Join our Discord: https://bit.ly/aibreakdown

Interested in sponsoring the show? nlw@breakdown.network