Today we celebrate: Towel Day

What Is the Purpose of the “Script” In Coding for Humanists

There are situations where a script is overkill: one-off investigations, quick checks, diagnosing a problem that often changes.

If you are a humanities scholar starting your adventure with AI-supported coding, you probably know this problem: you have an idea of what you want to achieve, but you get lost in the sequence of clicks and commands. This is precisely when a script comes in handy – clean and clear.

A script is nothing more than a recorded procedure, except that instead of being in a Word document, it is a set of commands that the computer can execute automatically. It’s not about programming for the sake of programming, but about not having to remember actions that always proceed the same way.

A script is not a clever solution but external memory.

My rule is simple: I create a script when I want to protect myself from my own mistakes. It may sound harsh, but it works.

In practice, in the trustedone.pl project, the most profitable scripts turned out to be those for monotonous but crucial tasks – such as where skipping one step causes chaos (e.g., implementing changes and clearing cache in various places). When I did it manually, I surprisingly often forgot. The script solved the problem: one run, fixed order, less stress.

When a Script Makes Sense – Version for Users, Not Programmers

You don’t need complicated decision matrices. Two questions are enough:

Do I do this often? If the task repeats regularly, automation starts to pay off. I follow the “rule of three”: if I do something for the third time, it’s no longer a single case but a process.

Is the error costly? If a mistake means an hour of fixing or a production embarrassment, the script works like checklists in aviation. No one pilots “by memory,” even on a good day.

When a Script Gets in the Way (and AI Is Better Then)

There are situations where a script is overkill: one-off investigations, quick checks, diagnosing a problem that often changes. In such cases, AI works better as a conversation partner: it suggests commands, helps compare results, suggests directions – without the need to create a tool that will be outdated in a week.

This is important because “dead scripts” tend to multiply. Suddenly, a directory fills up with files like check-v2-final2-last.sh, and no one knows which one is the right one. In my notes, I call this plainly: script sprawl – automation that starts to hinder more than manual task execution.

The Most Important Safety Rule Worth Knowing from the Start

Never put passwords in scripts.

This is not “IT paranoia.” It’s a security practice clearly described by OWASP: hardcoded passwords or secrets are a vulnerability because sooner or later such a secret will leak or become hard to manage.

If you need to automate an interactive process (e.g., logging in requiring a password), there are tools like Expect designed to control “talking” programs in the terminal. It works but requires caution – especially regarding secrets.

How I Understand It: “AI for Discovery, Script for Repetition”

AI works excellently when you’re looking for a solution. A script is indispensable when you already know the path and want to turn it into one click.

The most “human” definition I’ve kept for myself is this:

a script is a way not to have to negotiate with your own memory every day. And memory – as we know – always takes a vacation when you need it most.

Share: