The Existential Underbelly of AI Agents

I embarked on a spiritual experience the first time I was offered an AI agent. It claimed to have a personality, remember my preferences, and get better at anticipating my needs the more I used it.

For context, I spend most of my weekdays on a computer leading marketing for a software consultancy, and doing my best to remain human in the process. It’s like walking a tightrope.

The digital reality that I frequent is becoming increasingly dynamic and it’s not just because of AI chatbots. Autonomous AI agents are early but gaining traction fast. The Diffusion of Innovation was a theory I learned in college, but now I’m living it. We all are.

What are AI agents anyway? Depends on who you ask. Depends on your type of work. When it comes to coding, AI agents take action on multiple tasks to accomplish a goal instead of generating one single response. But when it comes to consumer technology, truly autonomous AI agents are yet to find their “market fit.” That was until OpenClaw. Eight months ago, an AI agent called OpenClaw shook the tech world. Here is how Every describes it:

What’s a Claw?

Your claw is a personal AI assistant that lives in whatever messaging app you choose—WhatsApp, Telegram, Discord, or SMS. You talk to it the same way you’d text a friend, without the need for special commands, coding, or even a new app installed. Under the hood, it’s powered by the same kind of AI behind tools like Claude and ChatGPT, running through OpenClaw —an open-source framework for personal AI assistants created by Peter Steinberger.

Here’s what’s most important: It has a personality, it operates 24/7, and it has the ability to change itself. It remembers your conversations, your preferences, and your ongoing projects. The more you use it, the better it gets at anticipating what you need.

[There is something peculiar about the way technology can be presented, and we should see it as such. Listen to the descriptors again: “Like you’d text a friend,” “It has a personality,” “The more you use it, the better it gets at anticipating what you need.”]

After I read that article, I began to impulsively daydream. The claims of a personalized creative partner that helped manage my work was compelling, especially in light of the challenges of self-management. I hadn’t even set up my Claw yet, but everything I encountered that day was alongside my imaginary AI Agent. It was mid-afternoon when I was processing out loud that I started assigning finite numbers to infinite ideas.

Technology has limits. I have limits. The total capacity of anything is 100%.

My thoughts continued…

Is 100% of this technology going to solve 100% of my needs? Of course not.

But it carried that kind of energy.

What does this technology solve for, anyway? Even if 100% of this technology offered outstanding pragmatic value, there’s still some percentage of my squishy human self that is unaddressed. Otherwise, I’m 100% robot.

That is when the fog lifted.

Somewhere between the promise of technology and reality, a gap existed, and that gap was my expectation. The bigger the expectations, the wider the chasm, the greater the distance you have to clear to get to reality, and the further you fall if you don’t make it across. I now deem this theory the Diffusion of Disillusion.

I encounter this chasm every single day, in tiny ways, when I prompt AI. Seen through the light of expectation, prompting is not unlike playing the slots. Imagine approaching a pulsing machine that has the potential to award you hundreds, if not thousands, of dollars. Some claim to play the game and remain completely detached. Most of us, however, feel the butterflies in our stomach. When it’s time to pull the lever, we do it with a certain finesse, and throw up prayer if we are feeling desperate. [I’ve actually never played a slot machine before. But I’m a competitive dice roller and the analogy is the same.]

When it comes to prompting, there’s even more control than pulling a lever or rolling dice—there is an art. With each word I type, I slide another dollar bill into the machine, and raise the stakes. Finally, I press enter. I bounce my knee. Check my phone. Notice the result. Collect my winnings.

Perhaps this is dramatized, but not imaginary. By definition, LLMs are a probabilistic technology. Unmet, unchecked expectations when prompting AI begin to wear down your cognition—perhaps more—even if it’s imperceptible. Some claim to remain completely detached. Most of us, however, feel the butterflies in our stomach.

This roller coaster is nothing more than our humanity at play. We are irrational beings with emotions and egos. We believe that we can influence, if not control, outcomes more than we actually can. For centuries, man has sought to understand the relationship between the will and its surroundings. When it comes to the behavior we can and can’t control, Augustine illustrates four chapters across humanity.