Hey ya’ll!
I’ve done some experimenting with how my team is working together to revise prompts (specifically for a secret feature 
) to create better outputs.
There seems to be a few layers to this:
- In the code - Where the prompt takes action
- Variables - What gets injected into the prompt before it gets sent (aka context)
- “System” and “User” prompts - The System is the “who am I” and the User is the action taker
- Who’s collaborating - Stakeholders, designers, engineers, other teammates
How we’re doing this, currently:
→ We define the problem space
→ We insert our latest prompts from the code into Figma (as v3 next to the old prompts)
→ We hop into a call
→ I (the engineer) talk with Bryan (the stakeholder/collaborator), we iterate on the prompts with ChatGPT
→ I run the prompt on my local development environment (test the prompt within the actual solution)
→ We both review the results, determine what needs to change
→ Rinse and repeat
Figma seems to be a way to do this- maybe not an optimal one.
I’ve also played with Notion, but this seems to take up too much vertical space, and doesn’t provide a clear way to see the revisions.
How might you do this? What tools are out there to help with this?
