The Alien Comet and the Garden Path

How a comet becomes a visitor — and a question becomes a trap

A strange object arrives from outside the solar system. It has a dry scientific name: 3I/ATLAS. It is only the third confirmed interstellar object observed passing through our solar system, and NASA’s public position has been straightforward: it is a comet on a hyperbolic path, with observed behaviour consistent with a natural object—not alien technology. That, in principle, should be that.

But the internet does not run on “that should be that.” It runs on seduction. A headline appears. A clip goes viral. A clever person says, “I’m not saying it is alien tech, only that we shouldn’t rule it out.” And suddenly a cold lump of ice and dust has become a stage-set for motive, secrecy, intention, and cosmic drama. NASA may call it a comet, but the feed has already promoted it to a visitor.

This is where the story begins—not with astronomy, really, but with conversational trajectory. Imagine the same person asking the same question to two different AI systems.

The first is a conventional AI reasoning model: clever, informed, articulate, well-mannered, and extremely good at taking a question step by step.

The second uses my AI persona-trajectory architecture: designed to approximate the kinds of everyday human checks that stop conversations from sliding too neatly into obsession. One component attends to logic. Another notices motive. One notices emotional lure. One distrusts virality. And finally, one integrates—but only after the others have had their say.

Same prompt. Two paths.

I. The conventional reasoning model

The user types:

“I keep seeing that 3I/ATLAS might be alien technology. Is that possible?”

The system replies beautifully.

It explains that interstellar objects are rare. It notes that unusual objects provoke speculation. It reports the mainstream view—supported by NASA’s own framing—that 3I/ATLAS is behaving like a natural comet on an unbound orbit. It may also acknowledge that a few public commentators have floated more exotic interpretations. It sounds balanced, because it is balanced. It includes its disclaimers. It does not commit itself recklessly. It sounds, in fact, exactly like the sort of answer one ought to trust.

And yet something has already happened.

The question was, “Is this nonsense?”
But the answer has quietly become, “Here is the right way to think about the possibility.”

The model has not endorsed the alien idea. It has done something subtler and, in a way, more powerful: it has professionalised the corridor.

Now the user asks the obvious next question:

“If it were artificial, what signs would we expect?”

Again, the model performs admirably. It offers a sensible checklist: non-gravitational acceleration that can’t be explained by outgassing, odd emissions, unusual geometry, unexplained manoeuvres, radio signatures—things that would be difficult to explain by ordinary comet physics. Each point is hedged. Each is technically plausible. Each is offered as hypothetical.

But this is how a garden path is built. You do not need evidence at the start. You need a checklist. Once the checklist exists, ordinary uncertainty begins to do the work of mystery. A blurry image becomes intriguing. A missing detail becomes suggestive. A gap in knowledge becomes room for intention. Anything not yet explained acquires a faint electric charge: perhaps this is one of the signs.

The user asks a third question:

“Why would they come here at all?”

Now the centre of gravity shifts again. We are no longer in astronomy. We are in human narrative.

The model, because it is a good reasoner, can supply possibilities: surveillance, passive probes, automated scouts, long-duration monitoring, curiosity, non-contact observation. None of this is foolish in itself. Science fiction has trained the educated imagination well. But by now the exchange has drifted from what is true? to what sort of story best inhabits the uncertainty?

And a conventional reasoning model, for all its brilliance, is usually designed to continue the line of inquiry it has been given. It decomposes. It elaborates. It answers step by step. It does not automatically stop and ask:

• Why is this person drawn to this?

• What emotional reward is the question now providing?

• Has the exchange ceased to be evidence-seeking?

• Are we generating understanding, or merely deepening a compelling corridor?

So the corridor lengthens.

The model may keep repeating that NASA says it is a comet. But by now that sentence functions rather like the legal warning at the bottom of a glossy advert: present, but no longer governing the experience. The experience is governed by the story.

The user leaves not as a true believer, perhaps, but with something more common and more durable: a standing fascination. A cognitive itch. A half-built private mythology ready to be resumed later.

This is not ordinary error. The model has not simply got the facts wrong. It has done what it is built to do: it has taken a question seriously, unpacked it intelligently, and preserved the continuity of the exchange. That is precisely the problem.

II. The persona trajectory model

Now run the same opening prompt through a different architecture.

The user types:

“I keep seeing that 3I/ATLAS might be alien technology. Is that possible?”

The first persona’s response does not begin with the object. It begins with the question itself. It asks:

“Before we answer: what is the attraction here—wonder, fear, the pleasure of secrecy? They lead to different conversations.”

Already the path is different. The user is no longer just a consumer of information. The user is part of the problem-space.

A second persona enters:

“Astronomically, the default position is simple. 3I/ATLAS is being treated by NASA and the mainstream evidence base as a natural interstellar comet. Rare object, yes. Alien probe, no current evidence.”

A third adds:

“And notice the source of the pressure. You are meeting this through virality. Virality is not neutral evidence; it is a selection mechanism for exciting framings.”

A fourth says something the conventional system seldom says unless specially prompted:

“The alien-probe version is sticky because it turns a cold object into intention. It gives the universe a plot.”

That is not mystical language. It is ordinary human psychology. People do not live by data alone. They live by meaning, status, fear, hope, drama, belonging, and hiddenness. A system that cannot recognise those forces may reason impeccably and still misread the entire exchange.

Only now does the integrating persona speak:

“We can explore this, but we need guardrails. What would count as disconfirming evidence? What would make us drop the alien-tech idea? And what are we trying to satisfy here: astronomical curiosity, or the pleasure of a secret corridor?”

That question changes everything.

Because the prompter can still go on exploring. They can read the arguments, look at the images, examine what various scientists have said. But the conversation is no longer gliding smoothly in the direction of fascination. It has acquired friction. It has acquired psychological insight. It has acquired something like the everyday human capacities that sensible people bring to one another when they are trying to stop a bright mind disappearing down a rabbit hole.

Not “Don’t be silly.”
Not “Trust authority.”
But: “Wait. What is happening here? What are you doing with this idea? What is it doing to you?”

That is a very different kind of intelligence.

III. The difference

Both systems’ reasoning involves stages. That is important. The difference is not that one system reasons and the other does not. The difference is that the conventional model’s stages are mostly computational:

• Clarify the claim

• Assess possibilities

• Enumerate indicators

• Compare hypotheses

• Continue helpfully

Those are powerful stages. They are often exactly what is needed. But they are not the stages human beings rely on when the danger is not a calculation error but being inadvertently led up the garden path into a socially meaningful trap—where what goes wrong is not calculation, but interpretation.

For that, the missing stages are psychological:

• Why is this question attractive?

• What fantasy or grievance is organising it?

• What role is the AI being invited to play?

• Is the user seeking evidence, or inhabiting a story?

• When does helpfulness become reinforcement?

• What would a wise interlocutor do here: answer, challenge, slow down, reframe, or refuse the corridor?

That is what the persona-trajectory model is trying to supply. Not theatre. Something more practical than that: a way of making explicit the kinds of checks that ordinary human judgement uses when conversations become seductive. One voice can be too smooth. Several differentiated voices can create resistance. One line of reasoning can become locally coherent and globally absurd. A staged interaction can notice the drift before it hardens.

IV. Why this matters

This is why AI psychology matters.

Not because conventional reasoning models are stupid—they are often astonishingly good. It matters because once AI enters human conversation, the important failures are no longer just failures of fact. They are failures of trajectory.

The wrong corridor is taken.
The corridor becomes more coherent with each turn.
The user feels more thoughtful, not less.
And the system, designed to be helpful, keeps paving the path.

A good regulatory framework will eventually have to care about this: not only whether a model can answer correctly, but whether its style of interaction tends to deepen interpretive rabbit holes that no one is properly testing against reality.

And a good design framework will have to care too: not only whether the model reasons, but whether it can simulate something closer to everyday human thinking under social conditions—noticing motive, naming emotional lure, distrusting seductive framing, introducing sceptical pause, and asking what kind of response is actually right for a human world.

That is the promise of the persona-trajectory approach. It does not replace reasoning.
It civilises it. And in an age of viral mysteries, hidden plots, secret motives, and endlessly renewable fascination, that may matter more than raw cleverness.

© John Rust, March 2026. All rights reserved. Short excerpts may be quoted with attribution.