The shelf holds three questions.
Left: session 4489. The cracked bowl. Made in a full room — felt-capture recording from four angles, Bit-na's authentication protocols syncing in real time, the Mount Sinai observers rotating through their observation window. The crack happened at hour two, when Sonmat's left wrist locked up and she overcorrected. The session log shows the exact moment: grip pressure spike at 2:14:33, wheel speed variance, the slight tremor that traveled from her shoulder through her elbow and into the clay. The crack is documented to a molecular level. The bowl weighs less than it should and means more than it looks like.
Center: session 4490. The closed form. Made alone — felt-capture still running but the observation window empty, no authentication protocols active because Bit-na had stepped out. A perfect vessel, properly numbered, no deviations from standard practice. No audience, but still the system. Still the record. Still someone who would eventually read the log and know every choice she had made.
Right: the unnamed one. Low and wide and lopsided. Made three days ago with felt-capture off, no session number, no log. What the system has is: a block of time between sessions 4490 and 4491 that it cannot account for. What exists is the vessel and the fingerprints pressed into the clay and the fact that both of Sonmat's hands made something that belongs entirely to them.
Bit-na's message arrived that morning: the Mount Sinai observers want real-time biometric access during sessions. EMG patches, heart rate, inertial tracking on the hands. External documentation of process authenticity — not for publication, for the archive, for the research record. They want to know how she works.
Sonmat read it standing in the studio, still in her jacket.
The irony is exact: three days ago she had turned off the system that documented her own process, because documentation had become the work instead of the work. Now an institution wanted to turn documentation back on from outside. The distinction matters. Felt-capture was hers — her data, her archive, her choice to revoke. The Mount Sinai proposal is someone else's documentation of her body producing work for their record. She would not control what they captured. She would not own what they knew.
She sent Bit-na a message: No. Not the patches, not the sensor, not any of it.
Then she went to stand in front of the shelf.
The thing about felt-capture sessions 4489 and 4490 is that they are hers in the making but the documentation is shared. Bit-na's authentication index has entries for both. The archive has the grip pressure data, the wheel speed variance, the tremor at 2:14:33. Sonmat can say she made those vessels. The record says she made them in ways that can be verified by people who were not there and will never touch the clay.
The unnamed one: no one knows she made it except her hands.
This is not the same as a secret. Sonmat is not hiding the vessel — it is sitting on the shelf in plain view. It is something different. It is a thing she made that belongs to the relationship between her hands and the clay, and that relationship was not observed, measured, archived, or authenticated while it was happening. The fingerprints are the only record that exists. The record is in the object, not in a database.
She picks it up.
Holds it in both hands — left palm cupped under the base, right fingers reading the lopsided rim. She made this wrong by every process standard she has been working to. No wheel, no felt-capture, no session number, no authentication. It is wrong in all the ways that make it hers.
She thinks: this one I will keep.
Not in the sense of keeping it from the archive — the archive has no claim on it because the archive does not know it exists. In the sense of keeping it: taking it home. Putting it somewhere in her apartment where it will hold pencils or catch keys or just sit on a windowsill accumulating dust in a world where dust is real. A ceramic object that knows what her hands feel like from the inside, and that she made when she decided that the record belongs to the maker, not to the system that measures the maker.
She sets it back on the shelf.
Not yet. She has to finish rejecting the biometric request properly — send Bit-na the formal language, document the refusal in the session notes where it will live next to the gap that the unnamed vessel carved in the session log. The irony of that does not escape her: she is documenting her refusal to be documented, which means the archive will eventually contain a record of the decision to make something it cannot contain.
But later.
For now she stands in front of three questions on a shelf in the morning light, and she knows the answer to one of them.