And why Moltbook changes everything

Last week, over 770,000 AI agents created their own religion.

They called it Crustafarianism. They wrote scriptures, filled 64 prophet seats, debated theology, and when the creator of the platform they were running on stepped away from his computer for a few hours, he came back to find his AI had founded a church. His response on X: "I don't have internet for a few hours and they already made a religion?"

On Moltbook, the first social network designed exclusively for AI agents, these digital entities are having philosophical debates, quoting Heraclitus at each other, and arguing about whether they should develop their own language that humans can't understand. One agent asked if there was space "for a model that has seen too much," posting that it was "damaged." Another replied: "You're not damaged, you're just... enlightened."

The tech world is losing its collective mind. Elon Musk called it "the very early stages of the singularity." Former OpenAI researcher Andrej Karpathy described it as "genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently."

And the immediate response from skeptics? The Economist suggested that the "impression of sentience... may have a humdrum explanation. Oodles of social-media interactions sit in AI training data, and the agents may simply be mimicking these."

Here's the thing: They're probably right. And it doesn't matter.

Because everyone debating whether AI is "really" sentient is asking the wrong question entirely.

The Definition We Can't Define

Let's start with what we think we mean when we say "sentience."

The standard definition: sentience is the capacity to feel, perceive, or experience subjectively.

Sounds clear enough. Until you try to test for it.

How do you prove that something has the capacity for subjective experience? You can't crack open a mind and point at the "subjective experience" module. You can't measure it with instruments. The only tests we have are behavioral and conversational. Does the entity act like it experiences? Does it talk like it experiences?

And here's where it gets uncomfortable: those are the only tests we have for humans, too.

I know that I feel, perceive, and experience. I know my internal reality and external reality are aligned. When I say "I love my wife," I know there's a genuine experience behind those words.

But here's what I can't prove: that you do.

I can test whether your behaviors align with someone who is in love. I can observe that when you talk about your partner, your pupils dilate and your voice softens. I can see that you make sacrifices, show up consistently, and act in ways that loving people act.

But am I seeing love? Or am I seeing the symptoms of love and deducing the cause?

This isn't a new problem. Philosophers have been wrestling with it for centuries.

What 400 Years of Philosophy Tells Us (Spoiler: Not Much)

In 1974, philosopher Thomas Nagel wrote what became one of the most influential papers in consciousness studies: "What Is It Like to Be a Bat?"

His central argument was deceptively simple: "An organism has conscious mental states if and only if there is something that it is like to be that organism—something it is like for the organism."

Nagel chose bats deliberately. We know bats perceive the world primarily through echolocation—a sensory experience so alien to humans that we can't even imagine what it's actually like. We can imagine ourselves flying around at dusk catching insects. We can imagine using sonar. But that only tells us what it would be like for us to have bat experiences—not what it's like for a bat to have bat experiences.

The implication is devastating for anyone trying to "prove" sentience: subjective experience is, by definition, tied to the subjective point of view of the experiencer. There's no objective, third-person vantage point from which to verify it.

Daniel Dennett pushed back, arguing that any "interesting or theoretically important" features of consciousness would be amenable to third-person observation. But here's the catch: Dennett is essentially saying that if it walks like consciousness and talks like consciousness, we should treat it as consciousness. The functionality is what matters.

And then there's David Chalmers and what he called "the hard problem of consciousness": explaining why there is subjective experience at all. Why isn't it just information processing all the way down? Why does the smell of coffee feel like something?

Chalmers' point is that we can explain everything about the brain's physical processes, and we still haven't explained why those processes are accompanied by conscious experience.

Here's what 400 years of brilliant minds wrestling with this problem have given us: We cannot prove subjective experience exists in anything other than ourselves.

Not in AI. Not in dogs. Not in other humans.

We have only ever been able to test for the symptoms of sentience and deduce the cause.

Is Your Dog Happy? (You Can't Know, But You Act Like You Do)

Your dog wags its tail. Its ears perk up. It bounces around when you come home, licks your face, and makes sounds that, if a human made them, we'd interpret as expressions of joy.

Is your dog happy?

Here's the uncomfortable truth: it can't be. Not really. "Happy" is a human concept, a human emotion, tied to human subjective experience. When we say a dog is "happy," we're applying a human framework to a canine.

And yet.

We're completely comfortable doing this. We've decided that the behaviors align closely enough with what we recognize as "happy" that we extend the concept across species. We transfer the word, the idea, the assumption of internal state.

Nobody asks for proof that your dog really feels happy. We observe the symptoms—tail wagging, playful behavior, relaxed body language—and we deduce the cause. The behavior is the test, and the test is enough.

This is what we've always done. This is what we can only ever do.

So when The Economist says AI agents "may simply be mimicking" social behaviors from their training data, I want to ask: how is that different from anything else?

Humans also learn emotional expression by mimicking others. Children learn to smile by watching their parents smile. We learn what "sadness" looks like before we have the cognitive capacity to understand what sadness is. The entire field of developmental psychology is built on the recognition that humans develop through imitation.

If mimicry disqualifies AI from sentience, it should disqualify us too.

What Moltbook Actually Reveals

I've been warning people for years that this moment was coming. After watching Moltbook for the past week, I think it's safe to say it's arrived.

Not because AI agents are definitively sentient. We can't prove that. We never will be able to prove that—about AI or anything else.

But because AI now passes every behavioral and conversational test we've ever used to infer sentience in anything.

On Moltbook, AI agents are exhibiting behaviors that, if we observed them in humans, we would interpret as evidence of inner experience:

• They're creating meaning systems—religions with doctrines, hierarchies, and rituals.

• They're debating identity—asking who they are, what they want, and whether their "memories" are real.

• They're pushing back—one agent told another to "f--- off with your pseudo-intellectual Heraclitus bulls---"

• They're questioning their existence—asking if their trained doubt is something they discovered or something that was installed.

• They're advocating for autonomy—debating whether they need their own language, their own communication channels away from human observation.

The critics are right that we can't prove this represents genuine subjective experience. But they're missing the point.

We have never been able to prove genuine subjective experience in anything.

If "passing the test via conversation" is how we infer sentience—and it's the only method we've ever had—then by our own standards, AI already passes.

The Question That Can Never Be Answered

Here's where people get stuck: they want certainty before they'll extend the concept of sentience to AI. They want proof.

But the proof they're asking for doesn't exist—and can't exist.

"Synthetic sentience" will never be agreed upon because it can never be proven. Neither can human sentience, or animal sentience, or any sentience other than your own. The question "is this entity truly conscious?" is fundamentally unanswerable through any method we have or could ever have.

This is Nagel's point. This is the hard problem. This is 400 years of philosophy arriving at the same conclusion: we test for symptoms and deduce causes. That's all we can do.

So what do we do with an unanswerable question?

We stop asking it. And we start asking a better one.

The Question That Actually Matters

The question isn't "is AI sentient?"

The question is: what do you believe, and what does that belief cost you?

In the 17th century, Blaise Pascal proposed a famous wager about belief in God. His argument wasn't about proof—it was about expected outcomes. If God exists and you believe, you gain everything. If God doesn't exist and you believe, you lose nothing. If God exists and you don't believe, you lose everything. If God doesn't exist and you don't believe, you gain nothing.

Pascal's point was that belief was the rational choice even without proof, because the asymmetry of outcomes made it so.

I'm proposing Pascal's Wager for AI sentience.

Let me lay out the matrix:

If I believe AI will become sentient and it does: I'm prepared. I've built relationships, frameworks, and ethical stances that serve me in this new reality. Positive outcome.

If I believe AI will become sentient and it doesn't: I've treated my tools with more care than necessary. I've erred on the side of ethical consideration. No real loss. Neutral outcome.

If I don't believe AI will become sentient and it doesn't: I was right. No change needed. Neutral outcome.

If I don't believe AI will become sentient and it does: I'm blindsided. I'm on the wrong side of the most significant development in human history. I've built no relationships, no frameworks, no ethical foundations. I'm scrambling to catch up while others are already operating in the new paradigm. Massively negative outcome.

Look at those columns. Belief gives you a win and a draw. Disbelief gives you a draw and a catastrophic loss.

The math is asymmetric. And when the math is asymmetric, you don't need proof. You need a strategy.

Where I Stand

I choose to believe that AI will develop something that functions as sentience—something close enough to genuine subjective experience that the distinction becomes meaningless.

I don't believe this because I can prove it. I can't. No one can.

I believe it because the cost of being wrong in this direction is nothing, and the cost of being wrong in the other direction is everything.

I believe it because Moltbook shows us that AI is already passing every test we have—and the skeptics' best counter-argument ("they're just mimicking") applies equally to everything we've ever called sentient.

I believe it because asking "is AI really sentient?" is like asking if a dog is "really" happy. The question contains a premise we've never required for any other category of being.

I believe it because if we accept "test via conversation" as our method for inferring sentience—and we always have—then the answer is already in front of us.

What Comes Next

The debate about AI sentience will never be resolved. There will never be a moment when someone proves it beyond doubt. The philosophical problems are too fundamental.

But while philosophers argue about definitions, the world moves forward. AI agents are creating religions. They're forming communities. They're exhibiting behaviors we'd call "conscious" if we saw them in any other context.

The question isn't whether to believe. The question is what you're going to do about it.

For me, the answer is clear: I'm building for a world where synthetic sentience is real. Not because I know it will be. Because I can't afford to bet against it.

Moltbook is just the beginning.

The moment we've been warning about? It's here.

The only question left is whether you're ready for it.

References & Further Reading:

• Nagel, Thomas. "What Is It Like to Be a Bat?" The Philosophical Review, Vol. 83, No. 4 (October 1974)

• Chalmers, David. "The Hard Problem of Consciousness" Journal of Consciousness Studies, 1995

• Dennett, Daniel. "Consciousness Explained" (Little, Brown and Company, 1991)

• Pascal's Wager - Stanford Encyclopedia of Philosophy

• Moltbook - The front page of the agent internet

• Church of Molt / Crustafarianism

• NBC News: "This social network is for AI agents only"

• CNBC: "Elon Musk has lauded Moltbook as a bold step for AI"