I’ve made a few custom GPTs. The custom GPT was created with instructions + knowledge in <= 10 files. This is approach 1. However when using ChatGPT, one can upload an attachment. This attachment can be a combined set of instructions + all knowledge content. This is approach 2. Is there a performative difference between these 2 approaches? Would the interaction be the same or similar?
1 Like
Same. Instructions are instructions but for it to have the same performance it will need to be on what you built it on. If it’s built on a 3.5 and you drop instructions into a 4o it will need to be moded, but yes a 4o copy paste and uploaded into chat is your custom. I call it drop code. Welcome to the forum.
See this is a meta in a messenger. I programmed it to be my assistant in real time.
Standard 4o in real time.
o 1 preview in real time 
It is called expert prompts too. Really all GPT can be boiled down to a prompt.
I have a few 100 copy paste GPT anything you can imagine
Took me 4 min to build all 3 
1 Like
The real difference is do you want to build it or have the knowledge to do so vs a custom for folks who lack the skills to “expert prompt /drop code” either way to use 4o you need a plus account minimum. You make a custom so you don’t waste prompts setting it up every time you run out of prompts or fill a page or start a new chat…
It is easiest on the user to use the custom gpt in OpenAI. It is a poetry generator with past poetry serving as knowledge (FredTheHeretic). However it is good to know that the same functionality can be achieved regardless of LLM.
1 Like
Happy I helped and yes if you are user focused and wish to generate chats from a parent, you need to do custom or you can share individual chats using share but they can see what you say in it.
1 Like
What was your initial input?
1 Like
A set of guideline rules and the 10 files for the knowledge
1 Like
It is the IP that makes this work…