Usages of ChatGPT that Won’t Compromise Your Company

There is no doubt that Large Language Models (LLMs) such as ChatGPT can be a huge boon for productivity.

As a software engineer, I’ve seen a huge benefit to my productivity by leveraging artificial intelligence tools. I’m able to find more concise answers to my questions, learn new skills at a rapid pace, and at times feel like I have my own personal pair programmer with me.

However, there are (rightfully so) data concerns with allowing the widespread use of ChatGPT in particular. The headline of Samsung employees leaking sensitive company data to OpenAI through ChatGPT underlines this point. I’ll step through some usages of ChatGPT that are safe and won’t compromise your company’s proprietary data.

Here’s the disclaimer:

Please use your judgment here and be sure to follow your company’s guidelines when it comes to using ChatGPT. I am not your compliance department and this is not advice on what is best for you and/or your company. Always check the output of ChatGPT before running code from it.

Note that in all of these prompts, I use placeholder variables that I then replace with values for my use case offline.

Scripting for data transformation & generation

ChatGPT is great for one-off, non-production-ready code that you can validate as correct manually. While some are using ChatGPT to format the data for them, I’ve been using ChatGPT to help me quickly whip up a script to transform the data myself.

Over time, I tend to build up a folder of data-wrangling scripts that have sporadic use. The great part about ChatGPT is that I no longer have to make the tradeoff of spending extra time to write a script vs manipulating the data manually because ChatGPT is so damn fast.

Example Prompt

Can you please write me a Python script that joins two CSV files on a common column named "bar"? Use the pandas library for brevity.

Shell scripting for developer experience

I am a lazy developer. I want to be able to Alt+Tab directly to my terminal and be able to get anywhere and everywhere. That means directly into containers such as my local database or API without even having to type docker container ls to find the container ID.

Example Prompts

Write a one-line bash script that grabs the container ID from a docker container, exec's into it, then provides the password parsed from a .env file in /foo/bar/location on my machine.

Write a short bash script that runs a JUnit test with the name “Testing Bar”. Set an environment variable API, with a default variable of “https://foobar” before running the test. Allow API to be overwritten as an optional parameter.

Debugging generic problems

I’m not a networking wiz by any means – while I grok the fundamentals, using ChatGPT to keep me on the right path speeds me up significantly.

Example Prompt

Outbound traffic is not routing through my VPN, how can I debug this from the command line?

Bonus Points: Follow up questions based on my specific setup, being careful to stay generic in your questions as you would posting a question to StackOverflow.

Scaffolding test cases

While I use IntelliJ’s Live Templates for this, I can see how some may find it useful. This is certainly not unique to ChatGPT, but it felt like an obvious inclusion.

Example Prompt

Write the scaffolding for 5 unit tests for me in Typescript in the Jest framework. Make sure to set up a beforeEach function, and label all of the functions as async. Use describe, add the boilerplate for assertions.

Bonus Tip: Disable pasting on ChatGPT’s interface

If you’ve made it this far, you may have noticed the commonality here is to not paste any code into ChatGPT. For those without self-control, here’s a handy Javascript snippet you can paste into the console (courtesy of ChatGPT itself) to help you resist the urge.

const textarea = document.querySelector('textarea'); 

textarea.addEventListener('paste', (event) => { 
  event.preventDefault(); 
  alert('Pasting is disabled for ChatGPT.'); 
});

---