loom-oi: A Hardware-Ready, Innate-Reflex Threat Detector Built on a Simulated Organoid's Own Response Magnitude

We report loom-oi: a working, hardware-ready system in which a simulated organoid's own firing intensity, driven by a single continuous physical quantity, is used by a reward-trained readout to reliably distinguish a calm scene from an approaching threat. The driving signal — frame-to-frame visual expansion rate — is not an arbitrary engineering choice. It is the same continuous quantity that triggers innate, unlearned defensive responses across virtually every visually-capable species tested, from insects to primates, via one of the most evolutionarily conserved circuits in the vertebrate brain (the optic tectum / superior colliculus). We build a synthetic organoid readout on the same principle: no new category is invented, no cognition is claimed. The readout learns only to interpret a magnitude the tissue already carries. Calibrating this system surfaced a genuine, non-obvious failure mode: a naive control condition's own measurement noise can exceed a real, weak version of the signal being detected, silently defeating any downstream classifier regardless of how well-tuned it is. We diagnose this precisely — down to per-class accuracy, not aggregate scores — and correct it at its source, not by patching the decoder. After correction, a systematic screen finds five independent random initializations that pass a three-stage qualification protocol (untrained separability, trained accuracy, and generalization to a held-out weak signal never seen in training), and, under an equal real-time budget, the organoid readout reaches 86% accuracy against a standard gradient-based digital model's 100% — a genuine, non-trivial comparison point, not the degenerate 100%/100% result an under-calibrated version of this same system produced. We report loom-oi as ready for its first real-hardware calibration step in a specific, bounded sense: the mechanism is verified, the failure modes that would otherwise consume paid session time have already been found and fixed in simulation, and a dedicated calibration script is built specifically to measure whether a live culture's own baseline activity — which cannot be set to zero the way a synthetic control's jitter parameter can — preserves the separation this system depends on. Source code, the full calibration history, and verification scripts: https://github.com/Metin1558/loom-oi