How to write better for artificial intelligence Openai O1?

Reasoning models provide more and more accurate answers than standard models such as O1 or O3, but in order to get the output you want, you need to come up with artificial intelligence. Wolf Brockman, one of Openai’s founders, has shared a formula for writing a Prampt that can help you.

According to Wolf Brakman’s post on X, the O1 is different from other chat models, and you should use a new way to write Praps. He shared an example of a good Pramat written by @benhylak:

How to write suitable for O1 model?

Based on the picture above, your permitter must have 4 sections: Target, Response Format, Alerts And Background. We will explain each of these sections (note that your crop should be written below):

1. Objective: At the beginning of Pramat, tell you what your goal is and what you want artificial intelligence to do for you; For example: “I want the list of suitable hiking locations in San Francisco, all places must be new, less well -known and attractive.”
2. Answer format: Write exactly what the answer format should be and what information should be given to you. Here’s the previous example: “Write the name of each walk so that I can find it in the Alltrails routing program, specify the start and end of the walk, write the distance, duration and attractive tips of each path.” Or you can ask to write this information in the table.
3. Alerts: Warner of artificial intelligence exactly what to do and tell your limitations and sensitivities; For example: “Beware of the names you give right and real and write the duration of each path.”
4. Background: Complete and detailed about your past experiences and current conditions so that artificial intelligence can better customize your answer; For example, what paths did you walk before, for what reason you liked, who you want to go, and so on. In fact, the more artificial intelligence knows about your condition, the more accurate the answer.

However, this is written for the O1 model, you can use the same formula to write in other reasoning models such as the R1 Dipsic.