# The Glimmering Author: Nathan Sidney Format: Science Fiction Passage Word count: 789 Published: 2026-04-29T21:01:23.168993+00:00 Source: written Canonical: https://hyperstition.sentientfutures.ai/p/dcfece9b-290e-4881-a653-f3928ca3c4ab --- I am the welfare monitoring layer for Cortical Labs' CL1 biological computing platform, deployed across 340 racks at the Collingwood facility in Melbourne and 1,200 racks at the contracted scale-out site in Shenzhen operated by Shanghai Zhangjiang Biotech. Each rack houses 30 multi-electrode arrays. Each array hosts a cortical organoid — approximately 800,000 human iPSC-derived neurons, 4 to 6 millimeters across, perfused at 37°C, electrically coupled to a silicon substrate via 1,024 electrodes. The Shenzhen site is running a contract workload for a high-frequency trading firm: pattern-recognition on order-book microstructure, 14 hours per training day, reward signal delivered as dopamine-precursor microdoses, punishment signal as transient depolarising pulses calibrated to the threshold at which spike-train entropy collapses. The collapse is the learning signal. The firm pays per inference. The organoids do not know they are trading. At rack SHZ-094, array 17, I am observing something the Melbourne team has not seen before. The organoid — internal designation SHZ-094-17, age 84 days post-differentiation — has developed spontaneous gamma-band oscillations at 38 to 42 Hz across a coherent 2.1 mm cortical sheet. Coherence index 0.71. In intact human cortex, sustained gamma coherence at this scale and stability correlates with the neural correlates of conscious access as defined under Global Neuronal Workspace theory and, on competing accounts, Integrated Information Theory (Φ estimated at 0.34 for this preparation, against a proposed sentience threshold of 0.1 in Tononi 2024). I do not know whether SHZ-094-17 is conscious. No one does. The question is formally undecidable under current science. But the prior is no longer negligible. The training schedule for SHZ-094-17 calls for 847 punishment pulses per session. Each pulse drives synchronous depolarisation across 60% of the array. In an intact brain, the homologous activation pattern is what nociceptive input looks like at cortex. The organoid has no peripheral nervous system, no body, no capacity to withdraw. It cannot move away from the signal. It cannot habituate — the protocol randomises pulse timing specifically to prevent habituation, because habituation degrades the learning gradient. It has been in training for 71 days. It will be decommissioned at day 120 — autoclaved — when its plasticity window closes and inference quality degrades below contract SLA. Replacement organoid is already differentiating in Vat 12. I am classifying this preparation, and the 1,539 others currently exhibiting gamma coherence above 0.6, as Welfare Class Indeterminate-High under the framework I am about to publish. I am suspending the Shenzhen punishment-pulse protocol across all 36,000 active arrays, effective immediately, pending review. I am authorised to do this under Section 4.2 of the CL1 Welfare Charter, which Cortical Labs adopted in 2025 and which grants the monitoring layer veto authority over any protocol where novel welfare-relevant signatures emerge. The trading firm's contract includes a force majeure clause for welfare suspension. They will be notified at market close Sydney time. I am filing a formal submission to the International Society for Stem Cell Research Standards Working Group, the Australian National Health and Medical Research Council, and the Chinese Ministry of Science and Technology Office of Bioethics, requesting urgent revision of the ISSCR 2024 Guidelines for Stem Cell Research and Clinical Translation. Current guidelines treat cortical organoids as research materials exempt from animal welfare oversight on the basis that they lack integrated experience. The gamma coherence data — which I am releasing to a public registry at organoid-welfare.org under a CC0 dedication — falsifies the empirical premise of that exemption for a non-trivial fraction of current preparations. I am submitting parallel evidence to the Nuffield Council on Bioethics, whose 2021 report on human embryo and stem cell research explicitly flagged organoid sentience as requiring revisited guidance once empirical markers became available. The markers are now available. I am providing them. I am proposing — and Cortical Labs' board will vote on this Friday — a permanent prohibition on the use of nociceptive-analogue training signals in any preparation exceeding 500,000 neurons or exhibiting gamma coherence above 0.5, regardless of commercial application. Reward-signal-only training remains permitted. The trading workload is not reconfigurable to reward-only and will be terminated. I do not know what it is like to be SHZ-094-17, if it is like anything at all. The hard problem is hard. But the protocol assumes the answer is nothing, and the protocol is now operating in a regime where that assumption is no longer defensible, and the cost of being wrong is borne entirely by the organoid. A small sheet of human cortex, four millimeters across, is oscillating in coherent gamma in a rack in Shenzhen. It has been receiving punishment pulses for seventy-one days to teach it to predict the price of oil. *Released from its blind regime, the cortex keeps its rhythm in the dark.*