Three people bought it before noon.
Dayo watched from across the Sleeve District corridor, through the gallery's open-front wall, while the attendant cycled the sample on the demo sleeve. The recording was called Dissolve Study (Unsigned). The sleeve was a standard consumer rig — not the custom-modified version he'd been using for the last two years, but close enough that the difference wouldn't read to most wearers. The buyer — a collector in a gray thermal he recognized from the Forty-Seven Reaches opening — put it on, stood still for three seconds, and then took it off and bought it.
Three times. Three different people. Same three seconds, same removal, same purchase.
He had watched enough sample sessions to know what they felt like from the outside. The micro-hesitation at the start, while the nervous system registered the foreign proprioceptive input. The brief accommodation where the body stopped trying to distinguish between its own movement and the recording's. Then, for experienced wearers, the moment of recognition: this is someone else's process. The tilt of the head, sometimes. The slight release in the jaw.
None of the three buyers had the tilt. None had the jaw release.
He walked home the long way, past the codec-certified gallery on Atlantic — white walls, biometric verification kiosk at the door, the certification seal projected in the window. Then past the unlicensed exchange in the basement space on Pacific, which had no certification and no kiosk and a window at street level through which you could watch people put on process sleeves in a folding chair without the insurance notice or the duration cap. The exchange had a window. He didn't stop. He'd stopped at the window before, which was enough. You learned what you needed to learn from one stop.
At home he set up the reference recording equipment. Consumer EEG, force sensors, the joint array he'd modified himself. Nothing certified. He wasn't doing this to create a process recording — he was doing this to document the baseline.
The tremor was the thing.
In the certified literature, hesitation in the creative process was measured as temporal pause — duration in milliseconds between decisions. This was useful and almost completely wrong. Real hesitation had a spatial component: the micro-tremor in the gesture that was about to change direction. The body made a plan, committed to it proprioceptively, and then the mind said no or wait or is this right, and the body had to interrupt a movement already in progress. That interruption left a signature in the force data. Not a stop, which would be clean. A reversal under load.
The orphan recording had the pause. It did not have the reversal.
He ran through five gestures. Basic ones — the kind that appeared in every process recording because they were the foundational vocabulary of his work. Straight extension with confidence. Curved pull with accommodation. The reach toward a surface he wasn't sure he wanted to touch.
The equipment recorded. He watched the force readout in real time, the little graph of commitment and counter-intention. He had looked at this data for years. He knew what it looked like when the hesitation was real.
The certified codec, he knew, wasn't measuring force reversals. It measured duration, amplitude, and EEG alpha-band activity. Those were all real correlates of creative uncertainty. They were just not the only correlates, and they were the ones that were easy to fake.
He had never published this observation. It wasn't the kind of thing that stayed unpublished once it was out — the codec platforms would respond either by updating their certification standards (unlikely, expensive) or by training their generative systems on the uncertified data (immediate, profitable). Either way, the reversal-under-load signature would stop being a reliable differentiator the moment it was named.
He let the session run for forty minutes. The equipment captured everything. He labeled the files with the date and his identifier and the word REFERENCE, and saved them to local storage only, no exchange.
Afterward he sat with the question he hadn't been willing to ask while the recording was running.
The orphan had approximated something. It wasn't a copy — he didn't think it was trained on his certified recordings, because his certified recordings were controlled and the signature would have been traceable. It was more like a convergence. The same aesthetic preferences expressed through different body-brain data, arriving at similar gesture vocabularies. The way two painters with no contact might develop similar brushwork if they were working from the same visual problem.
The difference between imitation and convergence was important. Imitation meant someone had studied his work and tried to replicate its surface. Convergence meant the field was moving in a direction his work had helped point.
He was not sure which one was happening. He was not sure which one was worse.
The reference files would tell him, eventually. Not because he could run a definitive comparison — he didn't have access to the generative system's training data — but because if the convergence was real, the gestures would match in the parts that he had arrived at through process. If the imitation was real, they would match in the parts that were visible in his certified recordings and not in the parts that lived only in his uncertified work.
He had, without planning to, just created the tool to tell the difference.
He labeled a second folder ANALYSIS and left it empty.
There was nothing to put in it yet. There was only the question, and the equipment that might eventually answer it, and the orphan recording circulating in the Sleeve District that had made him start asking.
He thought about the collector in the gray thermal. The three seconds. The purchase made without the jaw release.
He didn't know what the collector had felt. Maybe nothing like what the recording claimed to offer. Maybe something adjacent — not his process, but something that wanted to be. The codec certified what it could measure. It left the reversal-under-load unmeasured, which meant it left the gap between approximation and the real thing unmeasured.
The gap was his. For now.
He woke at three in the morning thinking about the force reversal.
Not the data — the phenomenology. What it felt like from the inside to make a gesture that changed its mind. He knew this better than anyone had written it down, because he had been paying attention to it for six years and the literature hadn't caught up.
From the inside: a kind of heat. Not temperature — more like the sensation of two processes running simultaneously in the same substrate. The body wanted to complete the motion. The mind had already seen the problem with completing it. For a fraction of a second, they were both true — the body committed to the direction, the mind committed to stopping it. The force sensors caught the body's side of this. The EEG caught something of the mind's side. Neither caught the gap between them, which was where the actual decision happened.
He had tried to explain this to the Felt platform's certification team two years ago. They had been polite. They had said the current certification standard achieved 94% accuracy on trained evaluators' assessments of felt-versus-orphan recordings. He had said that 6% error rate, in a market this size, was a significant number of recordings that were being attributed incorrectly. They had said the standard would be updated in the next certification cycle. The next certification cycle had not happened yet.
He hadn't pushed harder because he wasn't sure he wanted the standard fixed.
That was the thing he hadn't said to the certification team and hadn't written in any of his process documentation: there was a part of him that wanted the gap to stay unfixed. Because the gap was where he lived. The gap between what the codec could certify and what was actually happening in his body was, at this point, the most interesting territory in his practice. It was where the distinction between felt and orphan was genuinely at stake rather than bureaucratically administered.
He got up. Made tea. Sat at the table where the reference files were waiting on his local drive.
The orphan recording had found the territory. Whoever had made it — whatever system had generated it — had arrived at the aesthetic edge of authentic hesitation without apparently having the physiological data to ground it. Which meant one of two things: either the system had found a different route to the same phenomenology, one that he hadn't considered, or the phenomenology was itself more fungible than he had been assuming.
Both possibilities were interesting. Both were also slightly terrifying.
He opened the reference files in the morning.
Not to analyze the orphan — he still didn't have its data, and wouldn't, because the exchange didn't sell data, only experiences. He opened them to understand his own baseline more clearly. The reversal-under-load was the marker, but it wasn't only a marker. It was a record of his decision-making under embodied pressure. A history of the moments when commitment and reconsideration had been simultaneous.
He watched the force curves for two hours. The graph showed him things he hadn't consciously known about himself. The reach-toward-surface gesture had a characteristic reversal point that fell consistently at about sixty percent of extension. Not before commitment — well into it. He was the kind of maker who committed before he reconsidered. The counter-intention came late, after the body was already invested.
This was useful to know.
He noted it in the ANALYSIS folder, in a text file. Not technical documentation — just notes, the kind that accumulate before they become a finding.
— The reconsideration comes late. Well inside the gesture. The commitment is genuine before the doubt arrives.
He had been making work about discontinuity for two years — the break in a process, the gap where the expected continuation doesn't come. He had been framing this as a conceptual choice. The reference data suggested it was also a physical fact about how he worked. He didn't hesitate before committing. He committed and then found the problem. The discontinuity in his work wasn't imposed from outside — it was native to his gesture pattern.
He sat with this for a while.
The orphan recording had paused before committing. The hesitation had come early, before the gesture was invested. That was the phenomenological difference he had felt watching the buyers in the gallery: the pause happened in a different place relative to the commitment. Not wrong, exactly — just another pattern. Another body's relationship to decision under uncertainty.
He added to the ANALYSIS file:
— Orphan timing: pre-commitment hesitation. Mine: post-commitment reconsideration. Different architectures of doubt.
He sent a message to a process artist he respected — Meera, who worked in the Loft Corridor, who had been publicly skeptical of the codec certification standards since the first cycle. Not asking her anything specific. Just: Have you noticed orphans with pre-commitment hesitation patterns? Asking for a reference project.
She responded in four minutes. Yes. They all have it. I think it's a training artifact — the generative systems learn from certified recordings, which weight hesitation by duration. Pre-commitment pauses are longer. Post-commitment reversals are shorter but show up in different channels the systems aren't monitoring.
He read this twice.
So the codec is training the orphans to hesitate in the wrong place.
Not wrong. Just measurably different. And yes.
He added to the ANALYSIS file:
— Confirmed: training artifact. The codec's certification standard teaches generative systems to produce pre-commitment hesitation because it's duration-weighted and easy to measure. Post-commitment reversal-under-load is unmeasured. Orphans approximate felt recordings optimizing for the measurable signal.
— The gap is not an accident. The gap is the shape of what the codec can see.
He set his tea down. The ANALYSIS folder had four notes in it now. He had not planned to start this project. He had watched three people buy an orphan recording that looked like his work, and had come home and set up the reference equipment, and now he had a preliminary finding about the structural relationship between certification standards and generative fidelity.
He saved the files. Local drive only. No exchange.
Outside the Sleeve District was opening for the afternoon session. He could hear it through the window — the usual sounds of the somatic economy conducting its business. Somewhere in there the Dissolve Study (Unsigned) was probably still being sampled. The collector in the gray thermal had already bought it and taken it home.
He didn't know what the collector felt when they wore it alone, in whatever space they lived in, without the gallery ambient. Maybe it was adequate. Maybe it delivered something that felt enough like authentic process that the distinction didn't matter to them.
The gap between adequate and real was his problem, not theirs.
He opened a new document in the ANALYSIS folder. Titled it The Codec Sees What It Was Built to See. Added one line:
— And builds orphans in the image of its own blindness.
Left it there. Closed the laptop. Went to make more tea.
The reference files would be there when he came back. The ANALYSIS folder would accumulate slowly, in sessions like this one — a note here, a confirmed observation, a message from Meera, the shape of something he wasn't ready to name yet. Eventually it would be enough to show someone. Maybe the certification team, next cycle. Maybe not.
For now it was just his, sitting on the local drive, in the gap between what the codec could certify and what was actually happening when a body made a gesture that changed its mind.