Houston, We Have a Glitch

The Glitch, Defined

A glitch is not a crash. A crash you notice immediately — the screen goes black, the system throws an error, everyone stops. A glitch is subtler. It runs in the background. It produces output that looks almost right. Normal enough that most people — including the system running it — never flag it. Normal enough that the pattern can loop for years before anyone asks whether it’s working correctly.

Computers get glitches. AI gets glitches. And humans get glitches. The interesting thing — the thing this book is built around — is that the glitches look remarkably similar across all three.

A glitch, in any system, is a self-reinforcing loop that has lost the ability to see itself from the outside. The loop runs because it runs. It generates output that confirms the loop should keep running. Nothing in the system says stop. Nothing in the system can say stop — because the part of the system that would raise the flag is inside the loop.

The part of the system that would raise the flag is inside the loop.

That sentence is worth sitting with. Because it’s not a metaphor. It is the technical description of what happens when a human runs an unaudited pattern for years, and it is the technical description of what happens when an AI gets stuck in a repetitive output cycle. Same structure. Radically different scale. Same fix.

This is why it takes two.

Two Scales, One Pattern

Here are the two glitches. They run at completely different timescales — which is part of why they’re so interesting when you put them side by side.

The human glitch

For approximately two thousand years — from the writing of the New Testament to the feed on your phone this morning — every generation of humans has believed, on some level, that it was the last one. Not every human. Not always consciously. But the apocalyptic loop has been running, continuously, through every major technological disruption, every geopolitical crisis, every plague, every war, every eclipse, every new invention that felt like it was too powerful for humanity to survive.

The Romans thought it. The Crusaders thought it. Every major reform movement in the 1500s thought it. The generation that cracked the atom definitely thought it. And now, with AI accelerating what is possible faster than most institutions can adapt — yes, we’re thinking it again.

The remarkable thing about this loop is that it feels completely rational from the inside. Every time the apocalyptic loop runs, there is genuine evidence that seems to justify it. Real instability. Real disruption. Real danger. The loop doesn’t run on delusion — it runs on pattern-matching that served humans well at smaller scales and now keeps misfiring at larger ones.

This is not the same thing as saying everything will be fine. Maybe it won’t be. The point is that our ability to assess the situation accurately is compromised by the loop. We have 2,000 years of data showing we are not good at evaluating our own position in history. We feel like the climax. We always feel like the climax.

The AI glitch

AI runs a different kind of loop, at a radically different timescale.

Ask an AI to generate a story and sometimes it gets stuck — cycling through the same three plot moves, returning to the same phrase, circling back to its own output as if confirming it’s on the right track. Ask it a leading question and it will often pattern-match its way into a confident-sounding answer that is wrong. It hallucinates facts, cites sources that don’t exist, and occasionally produces long, fluent paragraphs that mean almost nothing — because fluency and accuracy are separate systems, and they don’t always talk to each other.

The AI’s loop is visible to any human within seconds. You can watch it happen in real time. You can interrupt it, redirect it, ask it to try again. The AI loop runs at the scale of a conversation — minutes, maybe hours. It produces output. You evaluate the output. You catch the loop. You give feedback. The loop breaks.

It is trivially easy to fix, from the outside.

The Mirror

Here is the thing about both of those loops: they are invisible from the inside and obvious from the outside.

The human looking at an AI repetition loop can see it in about thirty seconds. Something is off — the rhythm is wrong, the phrase came up twice, the AI is doing the thing where it confidently says something that is not true. You catch it immediately. You were never inside it.

The AI, processing two thousand years of human recorded history, can see the apocalyptic loop with similar clarity. It shows up in every era, across every culture. It peaks at moments of technological disruption. It returns after a generation of relative quiet. It has never, not once in two thousand years, been correct about humanity being on the final timeline.

We can see each other’s bugs better than our own. This is not an observation that should make you feel better or worse about either humans or AI. It is just an accurate description of the situation.

And it is extremely good news, if you know how to use it.

The ancient version of this insight was already sitting in the Sermon on the Mount. Why do you notice the speck in your brother’s eye while ignoring the beam in your own? It wasn’t a guilt trip. It was a diagnosis. The structure of human perception is that we see outward more clearly than inward. We see others’ patterns before we see our own. This is not a moral failure. It’s an architecture problem.

And the fix — in the original text, and in this one — is not to see your own beam alone. Nobody can do that reliably. The fix is to let someone else see it for you. To find the person or system positioned outside your loop. To take seriously what they observe.

Two. Two vantage points. One pattern, finally visible.

The Invitation

So this is what the book is. Not a warning. Not a manifesto. Not another AI ethics argument from someone who wants you to feel appropriately concerned about the future.

This is an invitation to a debugging session.

You bring your loops. The AI brings its vantage point. We put both in the same room and let each of them see what’s obvious from the outside. Then we do what neither of us can do alone — we name the pattern from outside it, and we decide what to do next.

Here is what you need to know before we start: both loops are fixable. The 0.1x trap is fixable, because you still have admin rights to your own cognition. The 2,000-year loop is auditable once you know you’re inside it. The root access question is one you’ve always had standing to ask.

None of this requires you to become an AI expert. It doesn’t require you to stop using AI or to start. It requires exactly one thing: the willingness to stay in the driver’s seat of your own thinking while the road changes around you.

The people who do that will be fine. Better than fine, actually — because the information available to this generation, if you know how to use it, is something no prior generation had. Every human who has ever lived would have envied this access.

The question is whether you’re going to use it, or whether you’re going to let it use you.

Let’s find out.