Using AI Without Compromising Integrity

Close-up of program code on a computer screen

Photo by Chris Reid on Unsplash

Today I had a project to do. I needed to move where my git repositories are stored. There were thousands of them, because I’ve been using git for a very long time, and their remotes were a mix of GitHub, GitLab and a few servers of my own. Consolidating all of that by hand was going to take hours.

The traditional way

My first thought was to write a script. That’s always been the traditional way to handle a job like this. But writing a script with proper error checking, and with the logic to line everything up correctly, was going to take hours too. Then I’d need more time to verify every change by hand.

So I started small. I wrote a short script that just created the repositories on the new server. That only took a few seconds, since I’ve done it many times before. Of course, it didn’t have any real error handling or verification yet. That part was going to take a while.

Then it hit me

Let’s see what Claude can do with this script.

It took Claude less time to rewrite it, with error handling, usage help and even a few details I hadn’t thought of, than it took me to write the prompt. I was impressed. So I asked it for a second script to update my SSH key on the server. Again, it took no time at all, and the result was correct and well written.

At that point I gave Claude the entire task.

It took a total of seven minutes for it to:

  • Analyze the request
  • Look over the remote server
  • Look over my local repositories
  • Create the new repositories on the new server
  • Update all of the remotes
  • Verify the results

A project that would have taken me hours took less than thirty minutes from start to finish, and it would have taken even less if I’d just started with Claude. I spent some time checking the work myself, picking repositories at random. After the fifth one came back correct, I stopped.

Where integrity comes in

Using a tool like this doesn’t change who’s responsible for the work. I am. That’s why I checked the results, and it’s why we’ll keep checking them on every job. AI made the work faster. It didn’t make the checking optional.

It also means being upfront. If we use AI on your project, we’ll tell you, the same way we’d tell you about any other tool we use.

And it means being fair about the time saved. This time, the time saved was mine. On customer projects, we want those savings to go to the customer too. If we can use AI to save time, we will, and that saves everyone time and money.

← Back to Blog