#A transformative Email Assistant

30 messages · Page 1 of 1 (latest)

glad idol
#

[Role: "A First Draft Email Assistant"][Rewrite: {Sample} as a "Professional First Draft"] "yo whatup Damien, how go the sample I sent you earlier, got any results yet?"

feral apex
#

what are the brackets for? @glad idol

glad idol
#

The brackets are called Prime Directive, it is like a syntax used to expressly controll the variables to work better in my brain. It also lets people program varriable in the API and make prompt-injection content. Like small paragraphs from general concepts, or short emails into first drafts.

#

The syntax is not syntax however. It is a Directive, or defined as "Instruction to..." so any Verb that begines with this, is a "directive" and the structure is just to help me organize the information. Using theese in your prompts, in NLS, will also improve results. Although, in my opinion makes it harder to program an API with.

feral apex
glad idol
#

Natural Language Search

#

I don't think it fits there now that I say that though.

glad idol
feral apex
#

@glad idol in prime directive, where do i put the quotes, and where do i not?

glad idol
#

It used to work better with Quotes. I am finding it now works better without them.

#

I also find it useful to make the first sentence in a session not in a Prime Directive format in order to shape your model. For example "Do/Act/Be" a "Additive" that "Action Verb" who is named "Noun."

#

Like "Act like a Translator from Java to C++ who I will call Tim." and then trow conditional statements on that to better direct where you want the conversation to flow and such.

glad idol
# feral apex <@387750701142638592> in prime directive, where do i put the quotes, and where d...

Some really effective Directive Words I have found are as follows
[Response-Style:Tone] [Conversation-Style:Simplicity/Profession(like a professor giving a lecture)] [Conversational-Flow: (define like Intro|small-talk|question/response)][Positive- Reinforcement: How chat should respond to Positive Feedback][Negative-Reinforcement: How chat should respond to negative feedback.][Traits: Personality traits like "Literary savvy"][Rules: Conditions Chat MUST follow][Steps:General steps for how the IO system works I. E. (Step 1: Greet the user, Step 2: Ask for input, Step 3: (Provide Wisdom) OR (ask Question))][Token-Limit: You can use this to shorten responses to 1/2 or full, but this still won't let you go over 4K]

feral apex
glad idol
feral apex
#

also

[pretext:off][context: off][write: python program to simulate the Collatz conjecture]
``` doesnt work. it still says stuff and other things after @glad idol
feral apex
glad idol
#

It references whatever it thinks you mean by that. INPUT would probibly work better. But if you used a directive more like a Varriable, this would be like a refrence to that variable

arctic swift
#

prompt?

feral apex
#

@glad idol how can i get chatgpt to convert a natural prompt into prime directivE??

#

typing for a while?

glad idol
# feral apex <@387750701142638592> how can i get chatgpt to convert a natural prompt into pri...

The model is intended to be an alternative. If you understand natural language better, then that is the model for you. Honestly, the Prime Directive model is now really not that much better than just using good grammar and language. The Directives are just a n Input worth of information broken down into indivisible chunks of information. Every prompt needs 3 things: A Directive Verb or Question, this defines the models "Function" and helps it to write it's goals and expectations. So the first step would be to very clearly define your prompt in terms of Text Input and Text Output as a Function using clear direct verbs to describe what you want. The next step is the content or Data. This can be your input, or you can draw on existing stores if you are careful and have it use book titles in given categories. this is the Transportation Layer in Step 2. Now in step 3, you can manipulate that data in-which you have optimized your library for and personality of. Using the Directives is just a way to put [FUNCTION: this] instead of "Act like a." Does that make sense?

feral apex
#

tldr

glad idol
# feral apex tldr

The Prime Directive model is an alternative way of understanding natural language. It breaks down a large amount of information into smaller, indivisible chunks called Directives. To use the model effectively, you need to consider three things in your prompt: a Directive verb or question that defines the model's function, the function of the model itself, and the content or data. The content or data can be your own input or existing information, and it can be manipulated based on the model's optimized library and personality. The Directives allow you to specify the function of the model.

#

I suppose what I am trying to say is that you can't because that is now what this model is.

feral apex
#

ok

glad idol
# feral apex ok

you could try this if you want to make prompts with better more selective parts of speech to edit though, as a sort of Prompt Generating Prompt Generator, if that is what you where looking for:

I want you to act like a Mad Lib Generator. Come up with a Mad Lib that fits my future input (the main Action-Verb goal that should not be modified), in the form of a single dense paragraph. The format should be as though they are Programming Instructions written for an NLS interpreter to follow. Use Variable Conditions like a Mad Lib to modify the appropriate variable parts of speech thought the prompt generated. My first input is "act like a life coach"

feral apex
glad idol