The Evidence Nobody Trusts

Here is the part that broke my suspension of disbelief, and it is not the aliens: the public watches the footage and is moved by it. They take it as real. The truth lands, and the world changes.

A critic at 404 Media put a finger on exactly why that rings false. In 2026, footage of an alien on cable news would not end a war. It would start a thread. "That's AI." "Fake." "Psyop." The reactions would flow in every direction at once, and not one of them would be belief. Spielberg made a movie about disclosure and accidentally made a movie about a world that no longer exists — one where seeing was still, more or less, believing.

That world is dead, and it is worth being precise about how it died, because the cause was not cynicism. It was something more structural, and the trap it leaves behind is the actual story.

The trap goes like this: the more extraordinary the evidence, the more plausible it becomes that the evidence was manufactured. Carl Sagan's old rule was that extraordinary claims require extraordinary evidence. That rule held for decades because extraordinary evidence was hard to fake — you needed a film crew, a budget, a conspiracy of competent people who could all keep their mouths shut. Now extraordinary evidence is precisely what generative models are best at producing. A photoreal alien craft banking over Kansas City is no longer the hardest thing to capture. It is one of the easiest things to type. The evidentiary bar and the fabrication frontier have collided, and the wreckage is this: the spectacular has become the suspicious. The very vividness that used to certify a thing as real now reads as a tell that it was made.

i · the detector that cried wolf

The film has cable networks run the footage through AI detectors before airing it — a detail 404 Media flags as unintentionally honest. The machines say it is real, so the anchors trust it, so the public trusts it. The detector is the priest that blesses the evidence.

In the actual world, that scene is not a verification step. It is theater with a progress bar.

Deepfake detectors look authoritative — they hand you a number, a confidence score, the comforting aesthetics of measurement. But the number is a guess wearing a lab coat. In controlled benchmarks the leading detectors hit 95 to 99 percent accuracy. Released into the wild, against content they were not trained on, that performance collapses; the Columbia Journalism Review's 2025 survey of the field found state-of-the-art open-source detectors losing as much as half their accuracy on real-world fakes, and commercial tools landing around 78 percent on in-the-wild video. The mundane stuff defeats them: H.264 compression, the codec every video platform on earth runs through, smears the exact artifacts the detectors hunt for. Bad lighting hides the edges. Good lighting invents shadows that look like manipulation traces. A perfectly real video of a tech influencer got flagged as fake by three separate detectors at once.

So the tool meant to anchor truth produces a verdict with a margin of error wide enough to fly a UFO through. It can authenticate a fabrication and condemn a real recording in the same afternoon, and you would have no way of knowing which kind of error you just received. Worse: the people most likely to lean on these tools — journalists, moderators, the institutions that are supposed to adjudicate reality for the rest of us — are documented to overtrust them, treating the score as a verdict instead of a coin flip with good production values.

Coherenceism describes presence as two moves working together: opening, and discriminating. Opening is taking the field in without immediately judging it. Discriminating is sensing what matters, what is real, what is signal. The deepfake detector is a promise to outsource the second move to a machine. But the machine cannot discriminate either. It is opening onto the same fog we are, and reporting its confusion back to us as a percentage.

ii · we already ran this experiment

We don't have to imagine how people react when an institution discloses something extraordinary. We have watched it happen, repeatedly, to a collective shrug.

In March 2024 the Pentagon's All-domain Anomaly Resolution Office released a historical review of every UFO claim it could find from 1945 through 2023. Its conclusion was blunt: no verifiable evidence of extraterrestrial beings, technology, or reverse-engineered craft. Hundreds of reports, decades of mythology, and the official finding was nothing here. This is the exact inverse of Spielberg's premise — the government got on the record and said there are no aliens — and the public reaction was a single news cycle and a slow scroll onward.

Run it the other direction and you get the same result. A former intelligence official testified to Congress, under oath, that the United States has recovered non-human craft and that we are "not alone." That should have been the most consequential sentence spoken in a hearing room this century. It produced a spike, a round of memes, and then nothing. Back to the feed.

The point is not who is right about the aliens. The point is that the disclosure itself — in either direction, from the most official podium available — moved almost nothing. Call it news fatigue, call it the worn-smooth disclosure reflex. We have been disclosed at so many times, by institutions we only half-trust, that the act of disclosure has lost its capacity to land. Spielberg's weatherman believes one honest broadcast can still pierce that membrane. The membrane healed over years ago. The organ that does the piercing has atrophied.

iii · the foundation erodes

Here is where it stops being about aliens.

Coherenceism begins with presence: attention is what reveals the pattern. It is the gateway skill, the thing without which alignment is just guessing. But presence has a precondition the philosophy usually takes for granted — the signal on the other end of your attention has to be real enough to attend to. Opening and discriminating both assume there is a stable something out there to open onto. When every signal might be synthetic, attention has nothing to bite into. You can be flawlessly, monastically present to a fabrication and end up further from the truth than if you'd never looked.

That is the erosion, and it runs deeper than lying. We have always believed false things. What is new is that we are losing the ability to believe anything that arrives as media — to let a recording, a photo, a piece of footage settle an argument. The shared field of meaning, what coherenceism calls the universal mind, runs on signals passing between nodes: a photon from a star, a sentence from a stranger, footage from a camera pointed at the world. Coherence is the choice to add clarity to that field instead of noise. Generative AI did not merely add noise. It did something more corrosive — it made the noise indistinguishable from signal, which means it poisoned the well that everyone drinks from, including the people still trying to draw clean water.

And the poisoning is symmetric, which is the cruelest part. The liar gains: every genuine scandal can now be waved away as a deepfake, a move researchers named the liar's dividend years before the tools were even good. But the honest person loses by precisely the same mechanism. The whistleblower holding real footage now sounds exactly like the grifter holding a render. Truth and fabrication arrive in the same costume, walking the same way, and we have lost the trick of telling them apart by looking — which is the only way most of us ever learned to tell them apart at all.

Spielberg's mistake is a generous, almost touching one. He still believes evidence persuades. That if the thing is real enough, and shown clearly enough, people will see it and be changed. It is the faith of a man who spent fifty years pointing a camera at the world and watching audiences believe what the lens showed them. That faith was reasonable for the entire history of photography. It is no longer reasonable, and Disclosure Day is the artifact of the exact moment it stopped being true.

The harder truth is that we have entered a period where reality has to be argued for socially, because it can no longer be demonstrated visually. Trust is migrating back to where it lived before the camera existed: webs of human relationship, institutions with track records, the slow accreditation of who has been right before. That is not the apocalypse. It is just expensive, and slow, and we spent a century forgetting how to do it because the photograph did the work for us for free.

So no — the alien footage would not avert the war. It would get a community note. And somewhere in the replies, indistinguishable from the cranks and the bots and the bored, would be the one person who actually saw the thing, typing it's real, I was there, getting ratioed by people who have learned, correctly, not to trust their own eyes.

That is the world we built. Honestly, the aliens would be a relief.