something i like doing lately:

write a long prompt (example below) of a detailed and open ended programming task and send it both to claude 3.5 and chatgtp

run both sets of code in parallell, and evaluate the output. compare.

this is the prompt rn

the output of both is exactly what makes this technique valuable

and then, for example, you can send back to claude the code that chatgtp replied and tell it:

"ok. i understand your reservations, so i wrote the code that you didnt want to write. please proceed with your suggestions building on top of this code"

father. working on /anky - new app coming out on the 20th of november

another useful prompt:

"ok. but that doesnt have any logs for me to read the console and understand what is happening. please add full detailed logs so that i can follow in every step. remember that i dont understand anything of what we are doing, but i honestly want to learn

help me do that"

pretty interesting first output tbh

(this is chatgtps code)

@thegoldenbright @king

Chilean musician and content creator mostly about Chilean folk music. 🇨🇱🧂Working on /chile 

Santiago, Santiago Metropolitan Region, Chile

I want to see when you Run the script on python!! How did that work?

too smart for my own good | living to learn, learning to live | "There are no accidents in the universe." ㅡ Wayne Dyer


in programming tasks, usually claude

the interesting part was that yesterday i sent both an 8 minute vomit of consciousness that i wrote through anky and their interpretation of it was almost the same

it was weird. v weird. and it made me think that maybe the LLM's will "converge" towards something

but idk

Wow thanks for the details! 
For the same prompt, which one performs better?
22 $DEGEN