Discussion about this post

User's avatar
FRED GRAVER's avatar

That is an EXCELLENT question. Ultimately, I think someone is going to have to design software that “runs” the virtual writers’ room. In the meantime:

- I use the tabs in google docs. After each response (from Claude, from ChatGPT, etc.) I enter that LLM’s draft into a tab (“from Claude” etc.) I read each of them, decide which one is the “foundation,” and then revise / copy / paste based on the input from the others.

Inevitably, in the course of doing this, I end up opening a tab labeled “new version,” that is a rethinking of the entire thing. I think that’s the best case for the virtual writers’ room — after reading multiple “takes,” you realize you have your own work to do. You do that and start all over again.

- ChatGPT’s desktop for Mac OS doesn’t have a quick click to upload google docs. I need to download the doc as a word file and upload that. UGH.

- Finally… in the end, I’m pretty much relying on my own memory and instinct to remember which version I like. At some point, you need to just go with a version and move on.

- BTW — chatGPT, true to its character as a really annoying recent USC film school grad — has offered to be my “organizer” for all of the LLMs. I told it to go get my car detailed.

Expand full comment
Morgan Beatty's avatar

a pain point with using multiples models simultaneously is versioning: how to keep each model up to date? Is it an exercise in meticulous copy/pasting and good labeling, or is there a more automated approach to feeding multiple models the same content? I know Claude can read updates to Google docs in real time. Can one google doc feed several AI's? Can one model be the organizer of versions of your work?

Expand full comment
4 more comments...

No posts