🌐 AI Agents Are Like Fish Tanks 🐠

The world of AI tools spins fast, and I find it hard to believe something is important to me just because a bunch of people are telling me so. While I focus on enjoying life as slow and as cool as a turtle, I accidentally discovered the "magic sauce" to AI agents from the most unexpected place: my aquarium. Let's embark on a quick adventure together that'll save you time, energy, and hopefully teach you a fresh perspective on AI tools you won't get anywhere else. Plus, maybe you'll even consider getting your own fish tank at the end of this. (The fish is optional.)

An AI Agent Is A Death Sludge (if you do it wrong)

On pen and paper, AI agents are pretty straightforward. You go to a url or you download a package and boom! You have an AI helper at your fingertips. Fish tanks are honestly exactly the same way. You're told to just get a fish, put it in water, and throw them both into a tank and boom! You have a slice of nature in your living room. That's all you need to succeed, right?

If you don't know already, if you just buy a fish and throw it into a tank in some tap water, you're soon going to be the proud owner of a brand new dead fish. Or death sludge, as I like to call it, versus the slice of the wild you were promised. With an AI agent it feels the exact same. You're told if you just start using your new AI buddy, you're suddenly going to achieve magic. Without additional context, instead you end up with an AI tool that may as well be that death sludge in your living room: you don't know what to do with it, and it's kind of hard to want to try again after that broken promise.

That magic is definitely possible to achieve and in way less steps than you'd think. Yes, for both the AI agent you don't know how to use and that fish tank you want to setup properly. You just need to focus on these three things.

Choose the Right Environment

To make sure you're using the best AI tool for your needs, you need to know the problem(s) you'd like to solve with it. When you know what problem to solve first, you'll look for a tool with the right features and context-learning you need. An aquarium works the same way! When you know what fish you're gonna get, you will know what features your tank needs and what environment you need to recreate. It can take a while to figure out, but it becomes way easier to review tool features you need and will save you the most time in the long-run.

At this point, I know I want a betta fish, which means I'll need a freshwater aquarium with a heater, filter, full-spectrum light, and some plants. Maybe some tankmates if I'm feeling spicy. For my AI agent, I know I want to start by getting help organizing hundreds of local graphic design files in my downloads folder, which means I'll need an agent that can look at set paths on my computer and edit/organize these files based on the instructions I present.

Know Your Parameters

Now that we know the problem we're trying to solve and what fish we're going to get, it's time to understand the parameters that will lead us to success. Another word for this can be goal posts, milestones to aim for. This will help you understand how to use the features provided to you, and how to start interacting with your AI agent best. There will be trial and error, but parameters make it a lot simpler.

For my betta fish, I learned she wants to live in water that's set between 75-80° Farenheit, in a tank with low flow to avoid damage to her delicate fins, set my light to the lowest setting, and add plants that will help the nitrogen cycle needed for a clean environment in this low-light setting. For my AI agent, I know I need something that can work on-machine vs in browser in order to review local files, to start with a single [test] folder, which I can use to develop instructions and start organizing.

Since we already knew what features we needed, it was easier to understand how to make use of them for our specific use case. I know where to point my AI agent to look at, and I know what to set my fish tank assets to.

The Engagement Cycle

We're already working on, or maybe have already solved the first problem we set out to do. With my files all neatly organized by date and type via a script, and a freshwater betta fish aquarium thriving in my care, what's left? You could honestly stop here, either until your next task or with no desire at the moment to continue. But...if you're looking to tackle bigger projects, create what hasn't been imagined, or take care of wild fish no one has tried, your engagement cycle is the main thing that will lead you to success.

For a fish tank, engaging consistently with it on a weekly basis is required in order to maintain its health, right? Otherwise you end up with death sludge. With your AI agent, while nothing will die from being left unattended, you won't tackle next steps or tasks you need to learn to solve without consistently using your AI tool. Whether you're trying to create an app with screenshot-driven development features, or you want to set up a saltwater aquarium with a whole octopus in it, it's the continued use and engagement that will get you there.

Thank you for going on this journey with me, and I hope you have a fresh new perspective and want to work with AI agents. While I go back to maintaining both of my new aquariums, I'll also be returning to the chalkboard to figure out how I can build an app with my AI agent of choice: Goose. If you want to learn more, check out the Goose repo and join our Discord community. Until next time!