Like many people, the solution to a coding problem I’ve struggled with suddenly appears when I try to explain it to someone or something. Sometimes even while sleeping.
This whole process of explaining to something else with the aim of figuring out the problem is known as rubber duck debugging.
A lot of people actually used to have physical ducks, mascots and maybe even the GitHub Octocat sitting right on their desks to come in clutch when they need it. Here’s a typical 100x developer’s table :D
In recent times however, the ducks have started talking back!!!!! 😲
AI tools like ChatGPT, GitHub Copilot (Chat) have slowly stepped in to replace the metaphorical duck. We are now able to not only explain whatever issues we are facing and hope to think of a solution in the process, but actually get suggestions and even outright solutions that we can then go back to try on our projects.
GitHub Copilot takes it an extra step, you can get code suggestions right in your IDE. And in complete honesty, that is still mind-blowing when I think back to the days I first started learning to code - thank you Dreamweaver.
But it can go even further.
What are AI Agents
You have probably already heard of AI agents somewhere in the winds. A simple way to think of an AI agent is as a personal assistant that is able to take your instructions, use a tool like ChatGPT to get more information, and then use this information to complete the task you assigned to it.
It can also ask you to review or present the final results to you when ready - all while you focus on other things.
These agents are often built on a LLM like GPTs, Gemini, etc but instead of just giving the responses back to you, they are able to autonomously act on these tasks for you.
Get it? 😂
Using AI Agents
As described, AI Agents can make attempts to accomplish any task you give to them - many are developer-only focused, some can help with refactoring code, migrating code bases, fixing CI failures, etc.
There are already many AI Agents out there - love this article from Ebony Louis that pointed out a couple you should check out.
I’ve been using Goose a lot recently - it’s open source so yaay. I am able to use Goose directly in my IDE within a project folder - when I need to, and basically tell it what I want to do.
I was able to build an application on livestream within an hour without writing a SINGLE line of code. As a developer, and developer advocate, I can’t begin to think of how many hours I could save with this.
The AI landscape is moving very fast. There are new releases shared every other day and I think we are up for quite an exciting time.
Recently I was able to share my thoughts on a State of the Frontend survey and it turns out about 75% of developers already use one form of AI or the other in their workflow. I believe the number is only going to go up and make these developers more efficient - embracing AI instead of being afraid of it.
I’m spending more of my time looking into how AI tools like this could help developers everywhere improve their workflow, and I can’t wait to share everything I learn with you all along the way.
Top comments (0)