TL;DR

• This is a few-shot style creativity test: give two new word definitions with examples, then have the model generate new sentences (or keep inventing words) following the same pattern.
• Good for: naming, brand copy, product terminology, and any "follow the example format" generation task.
• The key is making the pattern crystal clear: definition → example sentence, and constraining the output format.

Background

This prompt tests an LLM's ability to create new words and use them in sentences.

How to Apply

Think of it as minimal template learning:

• Give 1-3 examples showing the style and format you want
• Specify output: just one new word definition + one sentence, or N sets
• If you're using it for brand/product naming, add constraints (pronunciation, length, avoid sensitive words, avoid trademark conflicts, etc.)

How to Iterate

1. Lock the output schema: word / definition / example_sentence
2. Add constraints: root/syllable/semantic domain (e.g., must contain "-ify")
3. Add negative examples: which words NOT to generate (too similar to existing words, hard to pronounce, ambiguous)
4. Add self-check: have the model explain why the new word fits the pattern

Self-check Rubric

• Is the new word actually "new" (not an existing common word)?
• Is the definition clear, with no contradictions?
• Is the example sentence natural and does it demonstrate the meaning?
• Does it follow format and constraints (length, tone, domain)?

Practice

Exercise: migrate the "new words" pattern to your own business output format:

• Replace word with "feature name"
• Replace definition with "one-line product description"
• Replace example sentence with "in-app tooltip"

Prompt

A "whatpu" is a small, furry animal native to Tanzania. An example of a sentence that uses the word whatpu is:
We were traveling in Africa and we saw these very cute whatpus.

To do a "farduddle" means to jump up and down really fast. An example of a sentence that uses the word farduddle is:

Code / API

OpenAI (Python)

from openai import OpenAI

client = OpenAI()

response = client.chat.completions.create(
    model="gpt-4",
    messages=[
        {
            "role": "user",
            "content": "A \"whatpu\" is a small, furry animal native to Tanzania. An example of a sentence that uses the word whatpu is:\nWe were traveling in Africa and we saw these very cute whatpus.\n\nTo do a \"farduddle\" means to jump up and down really fast. An example of a sentence that uses the word farduddle is:",
        }
    ],
    temperature=1,
    max_tokens=256,
    top_p=1,
    frequency_penalty=0,
    presence_penalty=0,
)

Fireworks (Python)

import fireworks.client

fireworks.client.api_key = "<FIREWORKS_API_KEY>"

completion = fireworks.client.ChatCompletion.create(
    model="accounts/fireworks/models/mixtral-8x7b-instruct",
    messages=[
        {
            "role": "user",
            "content": "A \"whatpu\" is a small, furry animal native to Tanzania. An example of a sentence that uses the word whatpu is:\nWe were traveling in Africa and we saw these very cute whatpus.\n\nTo do a \"farduddle\" means to jump up and down really fast. An example of a sentence that uses the word farduddle is:",
        }
    ],
    stop=["<|im_start|>", "<|im_end|>", "<|endoftext|>"],
    stream=True,
    n=1,
    top_p=1,
    top_k=40,
    presence_penalty=0,
    frequency_penalty=0,
    prompt_truncate_len=1024,
    context_length_exceeded_behavior="truncate",
    temperature=0.9,
    max_tokens=4000,
)

Reference

• Sparks of Artificial General Intelligence: Early experiments with GPT-4 (13 April 2023)