[kingkongjaffa]

There's other criticism here about the writing quality (AI generated) but it's also quite out of date content-wise.

> Start with a basic prompt >Add Chain-of-Thought reasoning >Implement self-criticism >Add format constraints >Compare the quality and reliability of results

The 'techniques' are incredibly basic and Chain of thought isn't really relevant today considering there are models with COT 'baked in' i.e. the 'thinking mode' class of models.

[kingkongjaffa]

That being said I would love a up to date prompt engineering guide for getting the most out of the latest models in July 2025.

There doesn't seem to be too many incantations like there were before "think step by step".

The best thing I have found to use is meta prompting i.e. make a prompt designed to create prompts, and then use that resulting prompt to make specific targeted prompts.

You can then store these as custom GPTS in chatGPT or Claude projects.