4 Comments
User's avatar
John Brewton's avatar

Consistency comes from what you store, not what you prompt.

Daniel Ionescu's avatar

I find your playbook breakdowns very useful. Even if I use ChatGPT, they inspire my workflow and prompts.

The Synthesis's avatar

The cross-contamination problem is tool-agnostic. ChatGPT's Custom Instructions plus Projects get you 80% of the way there, the gap is that Skills load conditionally based on context while Projects load everything every time. If you're juggling multiple brands, the conditional loading is what prevents the 9pm Tuesday mixup.

Daniel Ionescu's avatar

Thanks for the clarification, it’s actually quite useful. And I’ve seen that you have options in ChatGPT to also restrict the memory only to a certain project.