How stable is a life surrounded by AI agents?

This is a reaction to Dan Koe's challenge of writing in a more authentic way in the age of AI.

I woke up yesterday.

First order of the day, I still had to go to the bathroom first thing. Some things never change.

Secondly, I didn't pull up Instagram anymore. The first thing that I do while seated on my throne is have a look at what my AI agents have been doing.

It's 2026, and the world feels like it's becoming more and more disconnected.

When I say the world is becoming more disconnected, I mean people becoming more disconnected. People don't talk to each other anymore. People talk to AI agents.

(I'm actually narrating this to an AI agent.)

I have a look on what I call the Agentic Command Center, which is a thing I built with agents. Four agents based on guidance by agents and after a discussion with my strategic agent.

What do you know? Agents everywhere.

(Yeah, most of them aren't really agents. It's just glorified chatboxes.)

I believe two years from this day, we're not even gonna think twice about the fact that we didn't have this just a short while ago. It would sound really weird if you would be talking to a friend saying, “Hey, I spent my evening chatting to artificial intelligence and I actually had fun.”

One thing that comes to my mind right now:

"I wanted to talk about the stable life or stability of life in the AI age, and I just realized the stability metric can be different for every single person." (I confess that this is a challenge of Dan Koe, which basically is about picking a question you don't have an answer to and freely exploring. So I'm just right now dumping my head on that topic.)

So what does stability, or maybe stability in an AI age, and maybe specifically for me, mean? What does it mean?

Here's how I imagine stability.

On the first of each month, I swipe my credit card and buy a subscription by my current AI provider of choice. Not gonna tell you which one it is now. I can tell you I just switched, as a matter of fact, today. And maybe that's also worth exploring a little bit later.

It tells me that I have access to AI on a specific subset of features based on artificial intelligence for the next 30 days or so.

The amount is pretty predictable. I would actually love to keep that at $20. There will probably be better and more expensive subscriptions, but one part of my intentional AI life is intentionally limiting the usage of AI to tasks that actually matter.

When I wake up in the morning, I don't have a look on what my agents did. I basically assume they did what they were supposed to do. They sorted out the problem, sometimes creating another one in the process. I think that's inevitable.

Stability also means that there is a little bit of a factor of instability, and I think I wouldn't be all too happy with the comfort of knowing that everything that I need can be completed by AI agents with 100% precision. I think it needs a little bit of a fail every now and then, just for the thrill and maybe for a realization that I'm alive.

(Maybe even for the agents to feel alive, apparently. What do you know?)

Stability also means that my life is progressively better every single day, and the people that I care about have a better life themselves as well, every single day. If they have their own metrics to that, well, so be it. For some, a better life is to just have a predictable day, getting a predictable salary, getting in a predictable number of challenges, ideally zero.

There's people like that, really. (When I think about it, it's the majority.)

So the core question, close to the end of May 2026:

"Are we as humanity in a stage where life with AI agents is stable?" "Do I have a feeling that my life is stable around, or surrounded by AI agents?"

Big fat nope.

Today's reality is actually pretty interesting.

We have technology that's existing for maybe seven to 10 years already. We're living in an age where that specific technology has been productized in a specific way.

Three and a half years ago, we were bluntly claiming that the world is racing with advancements of artificial intelligence.

But honestly, if you have a look at the technology itself, artificial intelligence hasn't made a lot of insane breakthroughs in the latest years.

What was insanely progressing is the rate at which we keep on building products and productifying artificial intelligence in nuanced ways.

My guess is, that it's the race makes life surrounded with AI agents quite unstable and unpredictable.

The only thing that you can really predict is the fact that the ecosystem is going to be changing every week.

So yeah, that's more of a confirmation rather than an argument towards stability.

Today it actually feels like we have the use cases figured out. It also feels like we have the prototypes figured out.

The things we absolutely don't have figured out are infrastructure, scaling, and business that will actually make AI companies profit. Also, intentions and sustainability - by faaar.

TLDR: it feels like we're throwing shit on the wall and trying to see what sticks.

From a business perspective, not much has been sticking, to be honest.

So what are the variables that need to stabilize?

Number one: the price.

Most AI companies are actually losing money every single day. Some of them are burning billions per year.

Investors are pouring humongous amounts of money into the ecosystem with the idea that this will eventually profit 10x or 100x, which is almost getting pretty unrealistic, to be honest, if I have a look at the scale of some AI operations.

We're becoming also, and that's number two, quite constrained with resources and infrastructure, which we're trying to figure out as we go.

And I'm gonna go back to this because this is one more aspect most people do not really see.

When I have a look at infrastructure, I see a series of overloaded clouds.

But in all honesty, cloud business is probably the business at the end today.

If you're letting people run their AI projects on your infrastructure, your utilization gets pretty high and with that, your margins are becoming pretty high as well.

Congratulations if you're one of them.

But the queue today isn't endless.

And even if we would produce a record number of servers, plug them into the cloud, the other constraint, and that's number three, is energy (and the energy grid).

We're not producing close enough energy.

We have old energy grids that were never really imagined to have a throughput like we will really need.

If we would have that energy, people would actually be looking for ways to make it much cheaper, which means coal and maybe a couple of riskier options, such as nuclear.

What's also pretty unknown today is reliability.

Not as in reliability of models. They have become pretty amazing, to be honest.

Models today are not even close to the early tests of GPT-2 or even GPT-3, which at that point blew my mind.

Now anyway, these couple of variables will make life in the AI space really unpredictable.

Which gets me to a question:

"Aren't we relying on AI agents too soon?"

Yeah, some people are. Some people are, indeed.

Is it a good risk to take?

Probably.

Probably still yes.

Let me give you my case.

Imagine that I wake up, do my early bathroom break, and have a look at my agentic center.

Instead of a nice report of “Hey, my AI agent did this, this, this, and this,” I would actually see:

“Oh, ChatGPT Codex is overloaded. Please try again in 177 minutes.”

And you're like, “Yeah, all right, probably overloaded, cool.”

Those 177 minutes have for sure passed, so you try again manually.

And it tells you:

“OpenAI ChatGPT is overloaded. Please try again in 213 minutes.”

So you're like, “OK, something's real. People are cooking.”

So you log in again in four hours to see:

“The models are temporarily rate limited. Please try again later.”

And suddenly, you don't know when later is.

And that later becomes the whole day of obsessively watching and trying to republish or restart your process.

Right now, it feels like I stated all my arguments, and I'd really love to maybe come to a conclusion.

So let's try to do a conclusion.

Is life stable for people in the AI age of AI agents?

Generally, no.

But yes for two specific groups of people.

Number one: people who have enough money and computational resources.

So as of today, a pretty decent computer with a high-end graphics card that can run acceleration and local models.

And number two: people who don't care about AI.

For the latter group, this is going to change. And I think this is going to change dramatically. This is going to become a skill that pays bills.

While we keep on saying that people can always come back to woodwork because AI is not able to do that, I saw some pretty amazing robots, and I saw some pretty amazing AI models powering robots. The technology is not yet there today, but it wasn't there three years ago for the iPhone either.

So yeah. People who have access to resources and are able to run high-end local models today are not gonna lose their edge.

They'll probably just keep on upgrading their rigs, run better models, actually learn how those models work, and how their daily tasks can be automated.

I do try to run my own local models, and there are some things that local AI makes quite predictable.

Among others, I know where my data sits, and I know what level of response I'm going to get with the same model every single day.

I'm pretty much fully aware of when the model changes, because I have to make that change myself.

For the others, still using third-party big providers, I think life's going to become harder.

They will become more dependent on that technology, even more than today.

Their thinking, their strategy, but also their level of thinking and talking is kind of going to become uniform.

AI and AI agents are by no means a bad thing, I think.

It's one of the best things that happened to humanity since the discovery of nuclear power, up until the point it became weaponized.

So I really do hope that I will still be living in an age where AI is not weaponized and predictable and really, really mine to operate.