Is Using Command Line AI an Option for Helping You Code?
Earlier today, I got curious and asked myself, “How well would using command line AI work for programming?”
I’ve used various IDEs to test how well their AI functions could work with creating programs of different types, but I’d not tried doing the same thing from the command line.
Today, I decided to change that.
Here’s what I discovered.
The Tools I Used
For this test, I worked on Pop!_OS Linux with the Ollama AI and a specific model that is geared toward coding. You don’t have to use Linux. As long as your OS works with Ollama, you’ll be good to go.
Speaking of which …
Installing Ollama
If you’re using either macOS or Windows, installing Ollama is as easy as downloading the installer file for your OS from the official Ollama site and running it.
With Linux, you simply have to run the following command:
Once Ollama has been installed, you’ll then need to download a programming-centric AI model. For me, I used qwen3-coder. To use that model, you have to first pull it with the command:
This model is roughly 15GB, so give it plenty of time to download.
When the download finishes, you then need to run the model with:
You should now find yourself at the Ollama prompt, ready to get to work.
Creating the First Program
What I decided to do was ask qwen3-coder to create a Python program that would accept user input for name, age, email and phone number and save the data to a file.
The specific query was:
It took qwen3-coder no time to start generating code. What it came up with was quite long and looked like this:
I copied the above code to a file named ai_input.py and ran it with the command:
To my surprise, it ran flawlessly.
Next, I decided to test it further by querying:
I was certain qwen3-coder would fail at this point. Sure, it might generate some code, but there’s no way it would get this right.
It took roughly the same amount of time for the model to spit out the new code, which looked like this:
That’s a lot of code, and surely it would error out, right?
It did.
However, the error was simply because the program couldn’t find tkinter. I installed it with:
I reran the code, and to my shock, the graphical user interface (GUI) opened (Figure 1).
Figure 1: My new Python GUI app works surprisingly well.
I could type input, save it, type more input, save it and keep going until I clicked Exit.
I decided to throw it for another look and ran another follow-up query:
Qwen3-coder went to work and spit out more code. I saved the output, ran it, and found that (once again) the app was successful.
Ruh roh. This time the program errored out with:
So, I ran another query informing qwen3-coder about the error to see if it could fix it. It went to work and believed it found the error. After it produced the new code: rinse, wash, repeat.
This time, the code was considerably longer (not sure why) and, according to qwen3-coder, this is what was fixed:
- Fixed the button creation syntax** — Added the missing closing parenthesis and proper styling.
- Added the missing
save_informationmethod** — This was causing the error since it was referenced but not defined. - Added proper file saving functionality** — The save button now actually saves data to a file.
- Fixed the lambda binding** — Made it properly formatted.
- Added proper error handling** — For file operations.
- Fixed the method call** — Made sure all method references are properly defined.
This time, the code ran fine. Although the background of the app wasn’t pink, the background for all the fields was.
Picking at nits.
The important thing here is that the Ollama command line AI tool was successful at what I’d asked it to do. Even with an error, I was able to do follow-up queries to resolve issues and even improve the code.
Does that mean you can stop learning the craft of programming and rely solely on AI? I wouldn’t. Because I know enough about Python, I am better capable of using the tool to either create something new or learn new concepts for the language.
Of course, your mileage may vary.
