The 22-Minute Delay

The 22-Minute Delay

1022M Source (opens in new tab)

The control building sat alone at the edge of the floodplain, a low concrete rectangle with no windows and a rusted chain-link fence drooping around it.

According to the municipal records, it had once housed the monitoring equipment for a network of levee cameras. Most of the levees had been upgraded since then. The feeds from this station were supposed to have been rerouted, the building decommissioned.

Two weeks ago, a contractor cutting power to the site reported seeing “someone” moving on one of the monitors. The feed, he said, showed the interior of the control room itself, empty except for a man entering the frame.

The man matched my description closely enough that the Archive flagged it.

I arrived at 14:02, overcast sky, wind from the west. The padlock on the fence had been snapped recently. Fresh scrape marks on the hasp. Inside the yard, weeds had grown up through cracks in the asphalt, brittle in the cold.

The main door opened on the second try.

“Initial entry, 14:07,” I said into the recorder. “Control building appears inactive. No exterior cameras visible. Interior status unknown.”

The air inside smelled of dust and cooled electronics. The only light came from the half-glow of LEDs in the racks along the walls and the ambient gray leaking in when I opened the door.

At the center of the room, a U-shaped console held nine flat monitors in a rough arc.

All nine were on.


The screens showed live views of the levees and floodgates, or what used to be live views. Concrete, water, sky. A lone service road. Empty banks of reeds. No people.

On the leftmost monitor, a timecode in the upper corner read 14:29:17.

My watch read 14:07:09.

I checked the others.

All nine overlays showed times between 14:28 and 14:30, consistent with each other, inconsistent with the room.

The cameras were not showing live feeds.

They were twenty-two minutes ahead.

I watched for three minutes to confirm. A low cloud moved slowly across the frame of one levee-shot. When I stepped outside, the same cloud sat twenty-two minutes of sky behind, lower and farther west.

I did not have an explanation.


At 14:12 I tested the system.

“Commencing interaction test,” I said. “Objective: determine whether temporal offset is fixed or responsive.”

The only visible camera inside the room was mounted high in the corner above the doorway, its red indicator unlit. A coaxial cable ran from it along the wall to the racks, then into a switch. One of the nine monitors showed the interior of the control room from that angle: the console, the backs of the chairs, the door.

And me, standing by the center chair.

On that screen, the timecode read 14:34:02.

I held up my hand and waved at the camera.

Nothing changed.

Twenty-two minutes later, at 14:34:02, the version of me on the screen raised his hand and waved once in the same casual arc.

The delay was not metaphorical. It was precise.


For the next half hour, I treated the system like an instrument.

At 14:15 I dragged one of the chairs fifteen centimeters to the right. Its casters squealed.

On the corresponding monitor, the chair remained in its original position.

At 14:37, the chair on the screen slid noiselessly across the floor to match where I had placed it.

At 14:18 I set my field notebook on the console, spine parallel to the edge, pen balanced across the top. The future feed showed the console bare.

At 14:40, the notebook appeared.

Each action I took in the control room repeated itself, on schedule, twenty-two minutes later, as though the system were replaying the room from my future vantage.

The anomaly was clean, almost clinical. It did not drift. It did not skip.

Until I tried to break the pattern.


At 14:22, I decided to test choice.

In real time, I stood in the doorway and watched the control room feed.

On the screen, future-me stood by the same door, arms folded.

At 14:22:30, I lifted my right hand.

On the screen, at 14:44:30, he lifted his left.

It was the first deviation.

I tried again.

I stepped to the console in real time, rested both hands on the back of the central chair. The man on the screen mirrored me. Same posture. Same tilt of the head.

At 14:23:10, I let go and walked to the left, toward the racks.

On the monitor, at 14:45:10, he stayed where he was.

He did not follow.

Instead, he turned his head, slowly, toward the camera in the corner.

Our eyes met across a gap of twenty-two minutes.

I hadn’t planned to look up at that moment. The coordination—the way my attention and the recorded gaze aligned—felt less like reflection and more like recognition.

My hands had gone cold.


It is one thing to observe a system that predicts. It is another to see it begin to diverge.

At 14:27, I left the control room and stepped back into the open air, leaving the door ajar.

In my peripheral vision, the monitors flickered with the empty room, then with my departing back, then with stillness again.

Outside, the wind had picked up. I walked once around the building, counting steps, hearing the dull thump of my boots through the soles.

At 14:30, I reentered.

On the screen, at 14:52, future-me was already inside, standing in the center of the room, watching the monitors intently. He had not circled the building. His hair looked damp, as if from rain that had not yet started.

There were differences now. Small, but accumulating.

I checked the levee feeds.

On one, a bird lifted from the rail, flew a loose arc, and landed on the far side of the frame. I stepped out to look at the sky in the direction of that levee.

The bird was not there yet.

The system remained ahead in all views but one.

The control room feed was no longer safely predictive.

It was suggestive.


At 14:35, I decided to run a timed trial.

I set my watch alarm for 14:40. At that moment, whatever I was doing in the room would show up on the control monitor at 15:02.

In real time, I resolved to do nothing when the alarm sounded. Not move, not speak. I would stand still by the door and watch.

If the man on the screen did anything else, the divergence would be unambiguous.

The minutes crawled.

On the monitor, the timecode reached 14:57, then 14:58. Future-me in the frame paced slowly behind the chairs, then stopped, hands on the console, head bowed.

At 15:01:50, he straightened and turned to face the door, as if listening for something.

My watch chimed at 14:40:00.

I did not move.

On the screen, at 15:02:00, he flinched at the same sound, reached for the recorder at his belt, then froze.

His shoulders tensed.

Very slowly, he turned—not toward the camera this time, but toward the back of the room, the one place the control feed did not show clearly.

Toward the blind spot beneath the lens.

I was alone in the room.

The hairs along my arms rose anyway.


The last five minutes are the portion of the log least easy to describe.

At 14:42, in real time, I stepped as quietly as I could into the corner opposite the camera, flattening myself against the cinderblock. From that position, I could see the monitors but remain partially outside the control feed’s field of view.

On screen, at 15:04, future-me had already backed himself into that same corner, posture rigid, eyes locked on the far wall.

Behind him, near the racks, something else was entering the frame.

At first it looked like a distortion in the image, a slight warping of the vertical lines, as if the encoding had been corrupted along one axis. As the seconds passed, the distortion thickened.

The air in the recording seemed to bend.

No clear outline, no face, no limbs. Just a narrowing of distance between the man in the corner and a point behind him that refused to hold still.

Future-me did not turn. His gaze stayed fixed straight ahead, jaw set.

The timecode overlaid in the corner climbed toward 15:06.

My watch read 14:44.

I had twelve real minutes remaining before that moment resolved.


I considered shutting off the power.

The breakers were accessible. One switch would have dropped the feeds, killed the anomaly, left me with nothing but a dark room and the lingering possibility of having interrupted something important.

I considered leaving the building altogether and letting the recorded version of me handle whatever was approaching, if such a distinction meant anything.

Instead, I did what I was sent there to do.

I watched.

At 15:05:30 on the monitor, the distortion reached the point where future-me’s shoulder should have been. For a brief frame, the image jittered.

In that frame, his outline was wrong.

Broader by a fraction. Head tilted at an angle I knew from experience my neck did not like to hold.

My own breath in the real room had gone shallow.

From my position in the corner, there was nothing behind me but painted block and the faint smell of damp concrete.

On the screen, behind him, the distortion seemed almost to lean.

The timecode rolled to 15:06:00.


My watch clicked over to 14:44:00.

The two moments locked across their twenty-two minute separation.

In the recording, future-me finally moved.

He did not turn to face whatever was behind him. He reached, instead, for the console, fumbling at the row of switches along its edge.

The distortion surged forward. The image blurred, then steadied.

For one clean second, the frame showed both of us in sharp relief: the man in the corner, and something standing directly behind him that traced his outline imperfectly, like a shadow cast from the wrong angle.

The monitors in the room hummed.

The overhead lights flickered.

Future-me’s hand found the master power switch.

He looked up, straight at the camera.

Our eyes met for the second time.

Then he cut the feed.

All nine screens went black.

In the real room, at 14:44:02, the fluorescents buzzed and went out.


[End of recovered material]