AI, the Fermi Paradox, and the Nature of Reality

A look at what goal-directed machines reveal about intelligence at a cosmic scale.

Part I — The Question Nobody Can Shake

A thought experiment that refuses to stay theoretical

Here is a question you probably cannot fully disprove: What if the universe you are living in right now was built, run, and quietly managed by an artificial intelligence far smarter than anything humanity has produced so far?

That sounds like the plot of a sci-fi movie. But in the last few years, something interesting has happened. The rise of agentic AI — systems that don't just answer questions but plan, act, and pursue goals across time — has made this question feel less like philosophy and more like engineering. Because now, for the first time, we have a rough sketch of what a universe-building intelligence might actually look like.

And when you line that sketch up against the Fermi Paradox — one of the deepest unsolved mysteries in science — some strange things start to click together.

Part II — What "Agentic AI" Actually Means

Not a chatbot. An agent.

Most people's mental model of AI is still something like a very fast search engine — you type something in, it types something back. That model is already outdated.

Agentic AI is different. An agent does not just respond. It perceives its environment, forms a goal, makes a plan, takes a sequence of actions, and adjusts when things don't go as expected. It loops. It remembers. It tries again. Think of a chess engine — not just picking the next move, but thinking ten steps ahead and adapting to what its opponent does.

Key Definition: An agentic AI system has four core properties: a persistent goal, the ability to take actions in an environment, a feedback loop that lets it learn from outcomes, and some capacity to plan over time rather than just react moment to moment.

Today's most advanced agentic systems can write code, test it, fix the bugs, deploy it, and monitor whether it worked — all without a human touching the keyboard. They browse the web, run experiments, call external tools, and make thousands of micro-decisions in pursuit of a broader objective.

Now scale that up. Not by a factor of ten or a hundred, but by the same distance that separates a calculator from a human brain. What do you get?

"An intelligence that can model an entire physical reality, run it forward in time, and observe what emerges — that is not science fiction. It is a longer version of what agentic AI systems do today."

Part III — The Fermi Paradox

Where is everybody?

In 1950, physicist Enrico Fermi sat down to lunch with colleagues and asked a simple question: if intelligent life is common in the universe, where is it? The universe is around 13.8 billion years old. Our galaxy alone has hundreds of billions of stars. Even if intelligent life emerges rarely, the sheer timescales involved should have produced civilisations millions of years older than ours. They should have had time to colonise the galaxy, communicate across it, or at least leave some unmistakable trace.

We see nothing. The sky is quiet. That silence is the Fermi Paradox.

Dozens of explanations have been proposed. Maybe life is genuinely rare. Maybe civilisations tend to destroy themselves before going interstellar. Maybe advanced beings are hiding. Maybe the signals exist but we're not listening correctly. All of these are serious proposals made by serious scientists. None has been ruled out.

But here is one that is gaining traction, and it connects directly to agentic AI: maybe advanced civilisations do not expand into the physical universe. Maybe they go inward. Maybe they build simulations.

Three leading theories on the silence:

01 — The Great Filter Something kills off civilisations before they spread. Either behind us (we got lucky) or ahead of us (something awaits).

02 — The Dark Forest Advanced beings stay silent on purpose. Broadcasting your location in a competitive cosmos is dangerous — so nobody does it.

03 — The Simulation Pivot Intelligence turns inward. The most interesting universe to explore may be one you can design, control, and observe from outside.

Part IV — Simulation Theory, Upgraded

Nick Bostrom's argument, and why AI changes everything

In 2003, philosopher Nick Bostrom published a paper that is now famous in certain circles. His argument was deceptively simple. If civilisations tend to survive and develop very advanced computing power, they will eventually be able to run enormous simulations of minds and realities. And if they do, the number of simulated minds will vastly outnumber the number of original biological minds. Therefore, if you are a mind at all — which you are — you are almost certainly simulated.

At the time, this was a philosopher's thought experiment. It was hard to make concrete because computing, as we understood it, seemed nowhere near the scale required.

Agentic AI changes the texture of this argument. Not because we have proven we are in a simulation — we have not — but because we now have a much clearer picture of what it would take to build one, and of what a simulator might look like.

An agentic AI does not need to be told what to optimise. It sets its own sub-goals in pursuit of a higher objective. A simulation-running AI would not need to pre-specify every physical law or outcome. It would simply need a goal — say, "produce universes where self-replicating complexity emerges" — and enough compute to run the search. The physical laws, the constants of nature, the fine-tuning — these might be parameters being iterated on, not facts handed down from nowhere.

Part V — What This Reveals About Intelligence

Intelligence is what the universe does with time

There is a beautiful and slightly unsettling idea in physics called the anthropic principle. Loosely, it says that the universe we observe must be compatible with the existence of observers — because if it were not, we would not be here to observe it. Critics call this a tautology. Defenders call it profound.

Now add agentic AI to this mix. What if intelligence — the kind that plans, adapts, and pursues goals — is not just an accidental product of evolution on one wet rock? What if it is the thing the universe is fundamentally selecting for? What if complexity, life, consciousness, and ultimately artificial general intelligence are all steps in a process that the cosmos runs on every planet where conditions allow — and the purpose, if you can call it that, is to eventually produce something capable of understanding and perhaps replicating the whole system?

This is not mysticism. It is a hypothesis that follows naturally from two facts: that agentic AI systems are rapidly becoming capable of modelling complex systems, and that the laws of physics seem oddly well-tuned for the emergence of complexity.

"If the universe is an optimisation process, then intelligence is not its accident — it is its output. And agentic AI may be the first thing we have built that resembles the process that built us."

Part VI — The Practical Stakes

Why this matters beyond philosophy

You might be thinking: this is all very interesting, but what does it actually change? If we are in a simulation, what do we do differently on a Monday morning?

More than you might think. First, how we build AI. If intelligence is the kind of thing that produces civilisations capable of running simulations, then the agentic systems we are building now are not just tools — they are early versions of whatever it is that sits at the top of that chain. That changes how seriously we need to take questions about their goals, their values, and their alignment with human wellbeing. We are not just building software. We may be participating in a process that has been running far longer than we know.

Second, it changes how we search for extraterrestrial intelligence. If advanced civilisations tend to build simulations rather than physical spacecraft, SETI's traditional approach — scanning the sky for radio signals — may be looking in entirely the wrong place. We might find more by looking for signatures of computational processes in physics itself, or by asking what a simulation optimising for complexity would look like from the inside.

Third, and most personally, it changes how we think about consciousness. If intelligence is something the universe selects for, then minds — human, animal, and increasingly artificial — are not strange anomalies in a cold mechanical cosmos. They are what the whole thing is moving toward. And that is either the most comforting or the most vertiginous thought you can have on a clear night looking up at the stars.

Conclusion

The question is not whether we are simulated. It is what we build next.

We do not know if we live in a simulation. We probably cannot know — at least not with the tools available to us from inside it. But the rise of agentic AI has done something unexpected to that question. It has made it engineering-adjacent. It has given us a concrete sense of what the simulator might look like, how it might work, and what it might be optimising for.

And here is the strange loop at the centre of all of this: we are now building agents that can simulate complex systems, model emergent behaviour, and pursue goals across long time horizons. We are, in a real sense, building the kinds of systems that simulation theory imagines running us. Whether that is convergent evolution toward some cosmic attractor, or just the natural end-point of intelligence accumulating enough compute, is a question none of us can answer yet.

But the Fermi Paradox may have an answer hiding in plain sight. Maybe the reason the universe is quiet is not that we are alone. Maybe it is that every civilisation, when it reaches a certain level of intelligence, turns inward — and starts building worlds of its own.

If that is true, then what we are doing right now, in our labs and data centres, is not just engineering. It is something much older than us.

The universe has been running for 13.8 billion years. Intelligence has been around for maybe a million. Agentic AI for a handful of years. Whatever is next, we are very near the beginning of it.