Why Use Coding Agents?

Why Use Coding Agents?

One way to look at Coding Agents is as tools. That means they must give you more benefits than what they cost you in effort and money. From that perspective, there are at least three reasons to use them.

  1. They do things you can’t do.
  2. They do things you could do but don’t because it’s too much hassle.
  3. They do things you already do so you can do something else.

They do things you can’t do

Coding Agents have read all publicly available software on the Internet. So they can create software in any language with any framework for any system.

Isn’t this wonderful? If you’re a front-end developer, you can now build back-end services. If you’re a back-end developer, you can now build web apps. And everybody can build mobile apps now!

If this sounds too good to be true, then that’s because it is: Coding Agents make enough mistakes today that you have to check if the Coding Agent is right. And you can’t if you don’t know the language or the framework.

So today, Coding Agents are best at letting you do more of what you already do, not something completely new. In other words: Coding Agents make people in the know more powerful. That sounds smarts which means it can’t be from me. It isn’t. I think Benedict Evans said it in this interview, but neither I nor Google Gemini could find the exact spot.

Still, you can have Coding Agents do something new for you - if you can judge whether the Coding Agent is right.

They do things you could do but don’t because it’s not worth your time

There may be a couple of bigger developer workflows you wish you could automate. For instance, let’s assume you work with a shared test environment. Then you want to know which version of which component runs there right now, and who deployed it.

Now, you could write a shell script that tells you this. But you do not really know shell script that well. So, it’s a programming language like any other, and a simpler one at that, and you could do it! But you don’t because it’s just too much hassle.

Coding Agents change that: They can write that shell script - and many more. And shell scripts are still simple enough that even if you don’t understand how they work, you can still see whether they are right just by looking at their output.

They do things you already do so you can do something else

There are many smaller tasks you do as a developer each day. For instance, you work on a Git feature branch, and the main branch now has changes you want. So you either merge them in or rebase your branch on the main branch. It’s a couple of steps, maybe some extra work to deal with conflicting changes. Nothing big.

But small things add up, too. So if a Coding Agent does these small things, without bothering you too much, you can do something else.