Hello friends,

Happy Friday and welcome to another edition of AI for the Rest of Us!

If you’ve been paying any attention to the news lately (or your social feeds), you’ve probably noticed everyone from Google to POTUS is talking about AI agents. Why? Because apparently 2025 is the year that AI evolves from communicating with you, to doing stuff for you. We’re talking about booking your flights, ordering your groceries, and finding those sneakers you’ve wanted (and making sure you get the best deal). It’s the year of agents rising.

This week, we’re starting a two-part series on these AI agents, and we’re doing it for three main reasons: (1) they’re mostly dominating the conversation right now, (2) they represent the next big shift in how AI will affect our daily lives, and (3) most importantly, they’re already here.

This week in Part I we’ll dig into what they are, how they work, and why the rest of us should care about them. For Part II in two weeks, we’ll talk about what to expect in the short and long term and explore the implications on culture and society. Buckle up because this topic might blow your mind a bit.

Here we go...

– Kyser

P.S. If you’re a new member and new to AI, I highly recommend you read Editions #2 and #3 - both of which answer the question, “What the heck is AI, anyway?”

Let’s start like we did in Edition #2 What is AI, anyway? with my most basic, no-duh definition of an AI agent and break it down:

What They Are
That’s heady stuff up there, so let me help crystallize it with a few metaphors:

Think of agents like having KITT or JARVIS on your iPhone instead of our old friend Siri. It’s a capable executive assistant that understands what you need and actually does the things you need – like respond to emails, schedule meetings, order groceries, and hey, one day drive your car.

You can also think of agents like having a virtual sous chef and admin in your kitchen as you prep for a dinner party. It would ask your guests if they have dietary restrictions, create a menu, tell you what ingredients you need, order all the groceries, design and print a menu, guide you step-by-step through cooking a demi-glace, and make sure your dinner party timing is perfectly coordinated.

Here’s one more for all you shoppers out there. Think of agents like upgrading from those “you might also like” recommendations to having a personal shopper who not only knows your style and size but also tracks sales, orders things it knows you’d like, and even sends your spouse a list of things to get you for your birthday, so hey, surprises can exist again in your marriage.

The key difference you’re probably already seeing between agents and the chatbots we’ve been using (a la ChatGPT, Gemini, and Claude): Agents don’t just respond to your questions or requests, they take action.

How They Work
Another way to understand agents is to see what makes them different from our current AI tools:

1. They’re Autonomous: Regular AI tools are like having a really smart friend who gives great advice but can’t actually do anything about it. Agents, on the other hand, can take that advice and run with it. They can make decisions and take actions without needing you to hold their hand every step of the way.

2. They’re Persistent: Unlike ChatGPT that forgets your conversation as soon as you close the window (unless you turn on that memory feature we talked about last edition), agents maintain awareness of their tasks and goals over time. They’re more like a personal assistant who keeps track of ongoing projects.

3. They’re Interactive: Agents can work with other tools and services. While Claude and ChatGPT can help you draft an email, an agent can actually send it. Same with calendars – an agent can actually schedule meetings on your behalf.

So hold up, you’re saying science fiction is actually becoming nonfiction? Yup.

What They Do
Here’s where things get interesting (and even more sci-fi-ey). Agents work by combining several AI technologies we’ve already talked about, then adding some special sauce:

First, they use Large Language Models (remember those from Edition #2?) to understand what you want and figure out how to do it. Think of this as the agent’s brain.

Then, they have what’s called a “planning system” that breaks down big tasks into smaller steps using specialized planning algorithms. If you tell an agent to plan my vacation, it might break it down like this: check your calendar for free dates, look up flights, research hotels, find activities, and create an itinerary. It’s doing this while knowing that you like window seats, don’t like staying in a room near the elevator, and love hiking.

Finally, and this is the game-changer, they have the ability to actually use other software and services. They might have access to your inbox, your calendar, or even your credit card (with your permission, of course 🤑).

Here’s a slightly more detailed example. You go to the website or app and give the agent a simple prompt: Please schedule a team lunch next week
The agent then gets to work:

• Checks everyone’s calendars for open blocks

• Looks up restaurants near your office

• Creates a poll to get everyone’s food preferences

• Makes the reservation

• Sends calendar invites to the team

• Books the Uber/s on the day of the lunch

• And because there’s a birthday to celebrate, it even orders a cookie cake from Great American Cookies® to be on the table when you arrive

This is all without you lifting a finger. At least that’s how it’s supposed to work. More on its trials and tribulations in Part II.

Why They Matter
Agents represent a shift from AI that helps us think to AI that helps us do. While tools like ChatGPT and Claude have so far been great for things like writing and analysis, agents are becoming great for taking action in the real world.

But this is all very new stuff. And we’re still in the early days – the very early days. The agents out there right now are like phones before the Blackberry and iPhone: basic and mostly one-dimensional. Not to mention, big tech companies have tried to do agents before and failed miserably, so it’s healthy to have a little apprehension. But it’s also healthy to have an understanding of them now – because things are moving fast and improving rapidly.

Which brings us to the big questions we'll tackle in Part II: What exactly can they do right now? What happens when these agents actually work consistently? When will I get KITT in my car? What does it mean for how we work, shop, and live? And perhaps most importantly: How do we prepare for them? Because whether we’re ready or not, agents are here. And this time, it looks like they’re sticking around.

What better way to learn about agents than to see them in action. Just last week, OpenAI (the company behind ChatGPT) launched their first version of an agent called Operator. It’s essentially a tool that uses the web to perform tasks for you, but that’s about it at this point. In fact, they’re saying, “It’s currently a research preview, meaning it has limitations and will evolve based on user feedback.” That’s another way of saying these things kinda work. Oh, and you can only use it now if you’re a Pro subscriber, which is a cool $200 per month. My recommendation: don’t do that, it’s not worth it. At least not right now.

We’re going rogue this week, and instead of sharing real-world examples of agents out there in the wild, we’re sharing a video that illustrates the concept in an entertaining way. It’s important to note that it doesn’t cover how AI agents will work in our day-to-day lives. We’ll get into that in Part II. This is more about showing you the concept of agents – with the bigger purpose of sparking your curiosity and lighting up your imagination (and maybe freaking you out a bit 😊).

Here’s the setup: a company put 1,000 AI agent characters into Minecraft to see how the agents would interact and organize themselves when left alone. What happened next was kinda bonkers. These characters didn’t just wander around aimlessly – they figured out trade, practiced religion, and engaged in democratic decision-making. They even created their own economy, formed a government, and wrote themselves a constitution. Which, let’s be honest, is the kinda stuff I wish my kids would do in there.

It’s Play Time

Newcomers [AI is new to me]

Special note to our new members: if you have never used ChatGPT and/or do not have access to it, then I highly recommend you read the Play Time section in Edition #1 and follow the steps laid out there before moving forward with this activity.

Since we’re talking about agents doing tasks for us, let’s start small and simulate what it's like working with an agent. Head over to ChatGPT and type in this simple prompt:

I want you to act as a shopping assistant that can actually buy things (even though you can’t right now). I need new running shoes. Show me how you would handle this if you could actually make the purchase. Break it down task by task and show me what you would do if you had the ability to do the shopping and purchasing. I wear size [enter size], I like [enter styles], and I have a budget of [enter budget range].

Now watch how it breaks down the task. Again, the goal isn’t to use it as an agent. The goal is to understand how an agent would think through and break down tasks. I’m hoping this gives you a taste of what’s coming when real agents can actually execute these plans.

Explorers [I’m comfortable with AI]

My original idea was to suggest you get a Pro subscription ($200/month) to test out Operator, but after playing with it, I couldn’t do it because it’s just not worth it right now. Don’t get me wrong, it’s a mind-blowing experience, but it’s not ready for the rest of us.

So for this week, I’m doing something random and answering a “reader mail” question I get asked fairly often: What AI tools do you use and for what? My list is short and sweet, and it isn’t mind-blowing or unique. It’s just what I tell people when they ask. Lastly, there’s more where this came from, so if you want more recos or have specific questions, hit me up.

• Claude for brainstorming or problem-solving

• ChatGPT for data analysis and visualization

• ChatGPT or Perplexity for any task you need web search. Note: be sure to check the sources because sometimes they pull from obscure websites.

• Claude for creative writing or really any writing

• ChatGPT for things around the house. E.g. Just yesterday, I took a picture of that confusing tag inside an article of clothing because I couldn’t tell from the wing-dings if I should dry it in the dryer or hang it up.

• Gemini Pro for pulling exact information out of long documents

• NotebookLM for querying / talking to files, articles, YouTube videos, etc.

• Fathom for meeting notes

• Deep Research for, yes, deep research on a topic. It sources info from dozens (sometimes over 100) sources and gives a thorough breakdown.

—

For those who want to dig deeper into agents, by all means, upgrade your subscription and play around with Operator. Or watch this longer explanation from the team at OpenAI:

This section is for premium members only.

And that wraps up Part I of our two-part series on AI Agents.

As always, please let me know what you thought and if you have suggestions on how we can improve AI for the Rest of Us. I’m always listening.

Until next time (and on Instagram between then)...