Yap Chuan Fu

Research Associate at The University of Manchester. I enjoy solving problems using computers and mathematics.

comment out to remove my picture-->


Other Posts


Summary notes from 'Pair Programming with a LLM'

Prompts for using LLM for programming - 5 min read

Published Sep 28, 2023

tldr; ‘everything’ you can do with LLM to help with your programming, please click on the navigation to the left to hop about.


With the advent of ChatGPT, large language model (LLM) or AI itself is now in heard and to an extent used by many people. There’s plenty of use cases, but deeplearning.ai have launched several short courses on generative AI; some of them were on how to use LLM for programming. Two courses that were of interest to me were:

1) Pair Programming with a LLM, this was built in collaboration with Google, which utilises PaLM LLM. Summarised in the current post.

2) ChatGPT Prompt Engineering for Developers this was built with OpenAI which uses ChatGPT LLM. Summarised in a different post over here.

Both courses were extremely interesting, and they taught me to write better prompts to make better use of LLM; in this case, to get better help with programming. This post will be compiling all the useful code and prompts as well as their respective use cases from Google’s course. For summary on the ChatGPT course, please click here.

Disclaimer: LLMs are not perfect, they suffer from ‘hallucination’ and make things up, so the code they generate should still be scrutinized and tested.


If interested in actually using PaLM LLM, please follow this link to get an API for yourself and access it. But I believe these prompts should would with any LLM, though obviously results may differ.

Further Notes on using the PaLM API (taken from the notes)

If you wanted to use the PaLM API on your own machine, you would first install the library:

!pip install -q google.generativeai

The optional flag -q installs “quietly” without printing out details of the installation.

Import and setting up model

The following is to import the model using the API key

import os
import google.generativeai as palm
from google.api_core import client_options as client_options_lib


# select model for text generation, which is what we need 
# there's other models available in PaLM
models = [m for m in palm.list_models() if 'generateText' in m.supported_generation_methods]
model_bison = models[0]

from google.api_core import retry
@retry.Retry() # decorator used to retry connection to google api if it fails
def generate_text(prompt, 
    "the temperature is 'randomness' of the LLM, higher more random"
    return palm.generate_text(prompt=prompt,

General Template for Everything

One thing highlight from the speaker is that ‘priming’ the LLM is really important, example below.

prompt_template = """



Your solution:

priming_text = "You are an expert at writing clear, concise, Python code."
question = "create a doubly linked list"
decorator = "Insert comments for each line of code."

prompt = prompt_template.format(priming=priming_text,

## Call the API to get the completion (which is the output from the LLM)
completion = generate_text(prompt)

The Programming Prompts

From here on, all ‘question’ space is the code you wish to get help with. The priming is fixed for the given use case.

The obvious thing left out is ‘please write XYZ code for me’ because this is about pair-programming.

Improve existing code

prompt_template = """
I don't think this code is the best way to do it in Python, can you help me?


Please explain, in detail, what you did to improve it.

completion = generate_text(
    prompt = prompt_template.format(question=question)

Alternative decorator:

Please explore multiple ways of solving the problem, and explain each.

Please explore multiple ways of solving the problem, and tell me which is the most Pythonic

Simplify code

# option 1
prompt_template = """
Can you please simplify this code for a linked list in Python?


Explain in detail what you did to modify it, and why.

# option 2
prompt_template = """
Can you please simplify this code for a linked list in Python? \n
You are an expert in Pythonic code.


Please comment each line in detail, \n
and explain in detail what you did to modify it, and why.

Write test case

This one is interesting as in can create tests for functions you do not originally have, which in turn is basically recommending how to improve your code.

NOTE: It may help to specify that you want the LLM to output “in code” to encourage it to write unit tests instead of just returning test cases in English.

prompt_template = """
Can you please create test cases in code for this Python code?


Explain in detail what these test cases are designed to achieve.

Make code more efficient

prompt_template = """
Can you please make this code more efficient?


Explain in detail what you changed and why.

Debug your code

prompt_template = """
Can you please help me to debug this code?


Explain in detail what you found and why it was a bug.

Explain complex code base

prompt_template = """
Can you please explain how this code works?


Use a lot of detail and make it as clear as possible.

Document a complex code base

prompt_template = """
Please write technical documentation for this code and \n
make it easy for a non swift developer to understand:


Output the results in markdown