OpenAI has released GPT best practices.
While experienced developers have figured out most of these, but this is a good comprehensive list.
👉 Write clear instructions:
GPTs can’t read your mind. If you want brief replies, ask for them. If you want expert-level writing, specify that. If you want a certain format, demonstrate it.
TIP: The less GPTs have to guess, the more likely you’ll get what you want
👉 Provide reference text
GPTs can invent fake answers, especially for obscure topics or citations. Providing reference text can help GPTs answer with fewer fabrications.
TIP: instruct GPTs to use citations from the reference text (or persona)
👉 Give GPTs time to “think”
GPTs make more reasoning errors when they answer right away, rather than taking time to work out an answer.
Asking for a chain of reasoning before an answer can help GPTs reason more reliably.
TIP: You want it to be a mentor or a consultant?
👉 Use external tools
You can compensate for the weaknesses of GPTs by feeding them the outputs of other tools.
For example, you can use a spell checker, a grammar checker, or a plagiarism detector to improve the quality of the outputs.
TIP: GPTs won’t solve world problem.
👉 Test changes systematically
You can use experiments and metrics to measure the impact of different strategies and tactics on your results. You can also compare different models and versions of GPTs to find the best one for your task.
TIP: Iterate. Iterate. Iterate
(via)
