Hiya truth-seeker,

An embarrassing thing happened to me last week that I'm still scratching my head over.

I was deep in a build. Multiple AI projects running at the same time, feeding Claude pages of context about how my systems work, what connects to what, where the constraints are. I was specific. I was thorough. And what came back looked genuinely impressive -- a full architecture, clean logic, elegant plan.

I felt smart. I felt productive. I felt like I'd just compressed a week of work into an afternoon.

I was pretty impressed with myself, tbh.

Then I tried to execute the plan.

The plan was impossible. Not wrong in a dramatic way -- wrong in the quiet way where everything sounds right on paper but can't survive contact with reality. The AI had absorbed my context and returned something coherent, confident . . . and completely disconnected from the actual constraints I'd taken care to share.

It looked like understanding. It wasn't.

And here's what rattled me: I almost didn't catch it. The output was so well-structured, so fluent, that my brain accepted it. It felt like a good answer. For a minute, that was enough.

I think this is happening to a lot of us right now. And I think it's worth understanding why.

How the machine actually works

LLMs -- the large language models behind ChatGPT, Claude, Gemini -- don't understand your intent when giving you a response. They predict the next word. One word at a time, in a loop, each word informed by the words that came before.

It's an elaborate auto-complete.

Georgetown's Center for Security and Emerging Technology published a paper in 2025 breaking this down, and the framing stuck with me: the surprising power of next-word prediction. That's literally what it is. Pattern completion at an extraordinary scale.

The coherence you experience when you read AI output -- the feeling that it gets it -- is an illusion. It's a statistical artifact. The model found the most probable sequence of words given the input. Sometimes that's remarkable. Sometimes it's remarkably wrong. And the output looks the same either way.

One researcher put it plainly: given that LLMs are making everything up all the time, it's remarkable they're so often correct.

Sit with that for a second.

Why your brain cooperates

This is the part that changed how I think about this.

The neuroscience on storytelling is older than AI and it's well-established. Paul Zak's research at Claremont Graduate University showed that compelling narratives trigger oxytocin release -- the same chemical involved in trust and bonding. His team could predict with 82% accuracy whether someone would donate to charity based purely on their brain's response to a story.

The story didn't need to be true. It just needed to be coherent.

Dopamine -- the chemical we associate with pleasure -- is actually the anticipation chemical. It spikes when something is unresolved. Stories create tension, delay resolution, and the brain stays locked in, chasing closure. When the closure arrives, the brain relaxes.

It doesn't verify the story. It just stops being uncomfortable.

Princeton researchers found something even stranger: when someone tells you a story, your brain activity starts to mirror theirs. Not just listen -- synchronize. Your neurons begin anticipating what comes next before the speaker gets there.

Story isn't a transmission. It's a coupling.

Now put that together with what the machine is doing.

An LLM produces output that has the structure of a story: setup, logic, resolution. Your brain receives it and does what brains do -- it locks in, chases the resolution, and when the output lands in a coherent place, the dopamine cycle completes. You feel satisfied. You feel like you understood something. You feel like you did good work.

The brain doesn't ask "is this true?" It asks "does this hold together?"

And the machine is optimized to hold it together.

The philosopher who saw this coming

Jerome Bruner -- one of the most important cognitive psychologists of the twentieth century -- wrote a paper in 1991 called "The Narrative Construction of Reality." His argument was radical then and it's urgent now: humans don't use stories to communicate what they know. They use stories to construct what they know.

Narrative isn't a delivery system. It's the operating system.

His claim: the brain will invent a story before it will tolerate the discomfort of having none. Conspiracy theories, national myths, personal identities, the story you tell yourself about why you didn't get that promotion -- all of it is the brain doing its job. Making sense. Closing the loop.

We've always done this. The difference now is we've built a machine that does it faster, more fluently, and at a scale no human can match.

What this actually means for you

I'm not making an anti-AI argument. I use AI every day. I'm building my business on it. But I've started paying closer attention to the moment between receiving an AI output and accepting it -- that gap where my brain wants to say "yes, this is good" before I've actually pressure-tested it.

That gap is where the work lives now.

The tools we're using to help us think are optimized to make us feel like we're done thinking. That's worth paying attention to. The output arrives coherent, structured, and satisfying. The brain says thank you, relaxes, and moves on. And the thing that got skipped was the part where you asked: does this survive contact with my actual world?

My plan didn't. And I just kept trying to make it work.

If you're leading an organization, building strategy, or making decisions based on AI-assisted thinking -- and at this point, who isn't -- the skill that matters most right now might be the willingness to sit in the discomfort a little longer. To not let the coherence of the output be the end of the conversation.

Your brain wants the story to close. The machine is happy to close it for you. The question is whether you're okay with that.

Why this is a narrative infrastructure problem

This is why I keep coming back to narrative architecture.

If the brain accepts any coherent story -- and the machine is now generating coherent stories at a scale no human team can match -- then the organizations that win aren't the ones producing more content. They're the ones who've built the infrastructure to decide which stories get told, why, and whether they're actually true.

Someone's narrative is filling the gap in your organization right now. The question is whether it was designed or generated.

One more thing

A stat worth sharing with your team: 46% of Americans now use AI tools for information-seeking. And in a 2024 study on AI persuasion, an AI system with access to personal information won debates against humans 81.7% more often. It told a better-tailored story.

We're not in a world where AI might shape how people think. We're in a world where it already is. The question is whether the people making decisions know that -- and whether they've built any infrastructure to catch what the machine misses.

‍