The tapes arrived in a gray plastic crate that still smelled faintly of floor wax and cold dust.
There were eighty-three cassettes in total, all labeled in the same tight, blocky handwriting: building code, camera number, date, start time, end time. A university campus, auxiliary storage facility. The crate came with a soft directive from the Archive: evaluate authenticity, assess whether reported anomalies warranted long-term retention.
The anomalies, according to campus security, were “visual artifacts probably from heat damage, but somebody upstairs wants another opinion.”
The crate was logged in at 09:58. I signed for it at 10:04. I began the first playback at 10:18.
The first tape that mattered was marked:
B-7 / CAM-03 / 2019-04-12 / 21:30–23:00
The footage was unremarkable at first. A fixed angle in a concrete corridor: storage cages, fluorescent hum just below the threshold of irritation, a wall clock in the far background. Boxes. Wire shelving. Occasional motion when students came to sign out equipment.
Nothing unusual, until a shift of grain in the far left corner of the frame resolved, slowly, into the suggestion of a figure.
It was positioned at the very edge, where the lens distortion widened faces and straight lines bowed slightly. Tall. Shoulders squared, back to the camera. No discernible features. The sort of blur that could be shadow, or damage, or a ghosting effect you only notice after staring too long.
I marked the timecode: 22:07:13.
I adjusted the tracking, then the brightness, then the color balance. The figure did not resolve into a trick of noise. It stayed constant as two students passed in front of the camera, signing a clipboard that sat on a table beneath it. Neither of them looked toward the corner. The figure remained motionless.
I rewound and watched again. The same.
The second tape came from a different camera.
B-7 / CAM-01 / 2019-04-12 / 21:30–23:00
Same corridor, different angle. Here the camera faced down the length of the hall toward a steel fire door. The table and clipboard were visible in the middle distance, distorted differently. The far left corner from the first tape was now a stretch of bare cinderblock halfway along the right edge.
The figure appeared at 22:07:13 on this tape as well.
Still in the extreme corner of the frame.
Still facing away.
Still motionless.
I tried to explain it to myself the way the campus technician had: a duplication issue, a burned-in overlay, some internal diagnostic pattern from the recording deck. But a control overlay would repeat across the full frame, or appear centered. This thing clung to the geometry of the room, always just barely within the field of view.
I checked the camera logs. CAM-01 and CAM-03 were different models installed three years apart. They used different firmware. Their only shared element was the power strip someone had plugged them into.
For the next six hours I lined up tapes and scrubbed through them. I did not take lunch. I did not speak.
The pattern held.
The figure appears on CAM-02 as a taller outline, the head scraping the top edge of the frame. On CAM-04 it is shorter, standing in front of a service hatch, shoulders sloped. On CAM-05 it seems too narrow, as if the body has been compressed laterally. Sometimes it is barely more than a darker band of vertical noise, but it is there: the same posture, the same orientation, occupying the same relative position no matter where the camera sits.
It never blocks anything important. It is superimposed just enough that I can see the room corner through it, faintly, as though the figure is not precisely opaque.
I made a grid of still frames at 22:07:13 from all cameras that covered the corridor that night. You could flip through them quickly and watch the figure flicker along the edge of the page like a badly printed animation.
The only constant was this: it never moved within any single tape.
If it was present at 22:07:13, it was present at 22:07:14, and 22:08, and 22:30, until the recording ended or someone turned off the lights. On some nights it simply wasn’t there at all. On others it appeared a few minutes earlier or later. But when it appeared, it was fixed, a stuck frame, an image burned onto the sensor.
That was the first assumption I had to let go of.
The second fell away when I played two tapes side by side.
I set up a dual playback: CAM-02 and CAM-03 from April 19th. On CAM-02, the figure stood at the far left, half obscured by a support column. On CAM-03, which looked down the corridor from the opposite direction, the corner where the figure should have been was empty.
Empty at 22:03:10. Empty at 22:05:00. Empty at 22:07:00.
It appeared on CAM-03 at 22:10:22. Exact.
Comparing the feeds frame by frame, there was no point at which the figure visibly walked from one position to the other. It simply failed to exist in one angle, then came into being already fully formed and static in another.
It had moved between cameras without ever moving on camera.
I checked the timecode sync and found a discrepancy of 0.47 seconds between the internal clocks, but the pattern remained. On several different nights, the figure disappeared from one field of view and appeared in another with no transitional frames. There were gaps when neither camera showed it at all, but its reappearance always recurred near the twenty-two-minute mark of an hour.
The security office had not reported any of this. They had simply written “visual artifacts” in the maintenance log and submitted a request for new cabling.
I started making my own notes.
2234-B / Observation 06 The figure’s apparent height varies between 1.55 m and 2.03 m depending on camera, but never inconsistently within a single feed.
Hypothesis (provisional): image is being composited at the level of recorded medium, not optical path.
Saying it on paper did not make it more reasonable.
Around midnight the building changed character. Daytime HVAC noise dropped off. The corridor lights dimmed by a fraction as occupancy sensors kicked in somewhere above. The only real light in the review room was the washed-out blue of paused footage.
At 00:09 I loaded a tape that had come with a handwritten note in the crate: “Security Incident, Cage 4B.”
B-7 / CAM-03 / 2020-02-06 / 20:00–22:30
Halfway through this tape, a student appears at the far end of the corridor carrying a heavy metallic case. He walks into frame at 21:51:09, fumbles with the key to his storage cage, and disappears from view behind a solid partition.
The figure is already present in the left corner of the frame when he arrives. It has been present, in fact, since 20:00:00, when the recording began. It has not shifted at all.
At 22:10:20, the student emerges, dragging the case behind him on a small cart. He passes in front of the camera and goes out of frame toward the exit. At 22:10:22, every light in the corridor flickers once and comes back slightly brighter.
On the next frame, the figure has rotated.
Not fully, not dramatically. Just enough that the line of its shoulders is no longer parallel with the wall. The posture reads as a person turning to look back over their shoulder at something behind them. The head is angled in a way that would, in a higher-resolution image, indicate an eye directed toward the camera.
The turn is subtle enough that if you were not primed to see it, you might miss it. I did not miss it. I rewound several times to verify that it was not a glitch in the tape or a bump in the playback mechanism. The edges of the shelving, the grain of the concrete, the wall clock — all remain perfectly stable between those two frames.
Only the figure changes.
I made another note.
2234-B / Observation 11 First instance of intra-tape positional change.
Note: coincides with corridor lighting disturbance and student departure.
Timestmp: 22:10:22.
I wrote the last colon heavier than intended and had to cross it out once.
I should have stopped there. That is a complete anomaly: a static artifact becoming dynamic at a precise timestamp, once, under clear conditions. It would have justified retention of the crate, a formal classification, one of the higher case codes.
Instead, I pulled one more tape at random.
B-7 / CAM-02 / 2020-02-06 / 20:00–22:30
The same corridor from the perpendicular angle. The same student. The same metallic case. The same time window. From this camera’s perspective, the figure should occupy the far right corner, just beyond a steel support column.
At 22:10:21, the student passes out of frame.
At 22:10:22, the tape drops four frames to snow and returns.
On the first clear frame after the dropout, the figure is gone from the corner.
Instead, for a single second, its outline appears closer to the camera, in the middle distance, where it should not be: partially overlapping the table and clipboard, bending perspective in a way the lens cannot quite resolve.
There are no more disturbances on that tape. The figure does not return.
The campus logs for that night list a complaint about “someone in the basement making me uncomfortable” phoned in by a student. No one was found when security checked the corridor. The guard wrote “everything clear” and closed the incident.
At 01:34, I shut down the monitors and re-labeled the crate for temporary retention. The overhead lights in the review room hummed at just above hearing for several minutes after power-down as the ballasts cooled. I stood there, listening, until the sound faded.
Before I left, I checked the wall clock out of habit. It read 10:22, though my watch said otherwise. The second hand twitched between marks, caught between one moment and the next.
I did not open the crate again that night.
[End of recovered material]