5 Comments

Great article 👍.

I use COSMOS for longer articles

- Context - I’m a….

- Outcome - I want to….info outcome

- Source - using these sources

- Mentor - using this type of style / voice etc

- Output - length type - summary etc bullets table

- Structure - ABT - Hegelian, PEE

Expand full comment
author

Oooh, I like that, I will have to give it a try. Just straight into ChatGPT? Or are you using another front end? Does the 'audience' go into Outcome? Ie. I want to teach content commercial real estate investors how to understand the impact of the current economic climate?

Expand full comment

Hi Sean

I’ve been using the framework like this (example for your field)

1. I’m an experienced commercial real estate journalist with the FT

2. I’d like you to write an Opinion piece for my column

3. Use these sources as background information (sources you plug in/reference)

4. Please use the Lex column for the style of writing and copy editing

5. Output length of 500-600 words using bullet point where required to help reader UX

6. Please structure piece in the ABT (And, But, Therefore) style used by Park Howell in his narrative gym

Expand full comment
author

Love it - I will definitely give that a shot in the future. I am often on the fence between requests structured as a 'one shot' like this vs. having a 'conversation' and getting to the point. I think both can work extremely well, and the one shot approach certainly has it's advantages (especially in a world with limited gpt4 model requests). However for people who are new to leveraging technology, I feel that treating the ChatGPT web interface like a chat client can have a lot of benefits in terms of teaching them about the capabilities of the platform.

One of the first posts I made here was pretty focused on the 'context' side of COSMOS you're describing: https://www.prompts.finance/p/crafting-stellar-chatgpt-prompts - and early on I found that little statement alone can be so powerful.

Maybe my next post on here will be diving more into this concept of learning by conversation and do a little COSMOS vs conversation comparison. No doubt you'll arrive at a similar spot, however, some people might need the journey to better learn how to take advantage of these tools for other use cases as well.

Expand full comment

Conversational defo works best for learning 👍 .

Like spreadsheets verbal formula work best by iteration - so I’ll often do a few tweaked versions and use best bits from each.

Exciting stuff Sean. Keep up the good work

Expand full comment