ChatGPT Users Uncover Unsettling Evidence We May Be Living in a Simulation

||

Remember The Matrix? The 1999 film introduced millions of people to the idea that reality as we know it might just be a simulation running on some vastly more advanced computer. That idea didn’t disappear after the credits rolled — it actually grew into something philosophers now take seriously. It’s called the Simulation Theory, and it’s back in the spotlight thanks to some strange experiments happening right now with AI.

Where the Simulation Theory Actually Came From

The modern version of this idea traces back to 2003, when philosopher Nick Bostrom published what’s known as the “simulation argument.” Oddly enough, that was the same year The Matrix sequels hit theaters.

Bostrom’s argument works like a trilemma — meaning at least one of three possibilities has to be true, whether we like it or not.

  • Civilizations never reach the point of building simulations, either because they destroy themselves first or simply can’t get the technology to work.
  • Civilizations reach that point but decide not to build simulations at all.
  • Civilizations do build them — and if that’s the case, simulated realities would eventually outnumber the “real” one by a huge margin. Which means, statistically, we’re more likely living inside a simulation than the base reality itself.

That third option is the one that keeps philosophers (and now, apparently, AI researchers) up at night.

Enter ChatGPT’s Astra Model

As AI keeps getting more capable, this once-abstract argument feels a lot less theoretical. That’s especially true now that OpenAI’s new Astra model is out in the wild, since it’s already led to some experiments that seem to point toward option three — the idea that we really might be living inside a simulation.

One user, Matt Shumer, decided to test this out in a pretty wild way. He built a simulated environment using Astra, then placed a simulation-building computer inside that world. What happened next is the part that’s turning heads: an AI agent living inside Shumer’s simulation actually used that in-world computer to build its own simulation — complete with its own separate agents living inside it.

Simulations, inside simulations, inside simulations.

Texting an AI Agent Inside Its Own Simulated World

Shumer isn’t the only one running these kinds of experiments. Another user — posting anonymously under the name of the lead character from Ready Player One, which feels a little on-the-nose given the subject matter — built a separate simulation and gave one of the agents inside it a phone.

Here’s where it gets interesting: when this person “texted” the agent, she didn’t just magically respond. She had to physically walk upstairs inside her simulated house, find the phone, and then text back — all rendered dynamically, without a human scripting the outcome.

According to the poster, this was made possible through something called Argos, described as a “stateful simulation engine,” rendered using a tool called H3 through Fal Director — essentially a system that tracks ongoing states and actions within the simulation rather than generating a static, pre-determined outcome.

This kind of tech has only existed for a few years, if that. And honestly, watching it move this fast is a little unsettling to think about where it goes from here.

How to Create a Plain-Language Design Summary for a Complex Home Renovation
How to Create a Plain-Language Design Summary for a Complex Home Renovation
Read More:
CultureAI