organoid-oi v3.1: A Biologically Grounded Closed-Loop Framework for Organoid Intelligence Simulation — and a Self-Audit Retracting Its Central Learning Claim

CORRECTION NOTICE — This version supersedes v1.0 (doi:10.5281/zenodo.20569348). Two claims made in v1.0 are retracted here: (1) the framework passed 45/47 synthetic validation tests, not 47/47; (2) the freeze-weights test does NOT confirm that STDP is the source of learning, and STDP is shown here to make no measurable contribution to categorical accuracy. The framework code is unchanged in its mechanisms; what changed is what can honestly be claimed about it. We present organoid-oi v3.1, an open-source Python framework for simulating closed-loop visual categorical learning in a synthetic brain organoid, together with a full self-audit conducted before any hardware deployment. The system implements Three-Factor Spike-Timing-Dependent Plasticity (STDP) with biologically realistic reward and penalty delivery: reward is a 10 Hz sub-threshold theta depolarization current producing natural long-term potentiation (LTP); penalty is a 200 Hz noise current producing long-term depression (LTD). A local Hebbian readout layer with weight decay decodes population activity. Every mechanism has a named biological counterpart — no backpropagation, no external weight forcing, no mathematical shortcuts. The audit reverses the central claim of the previous version. Under paired control (identical seed, identical stimuli, STDP enabled versus disabled), categorical accuracy is identical to the digit across every condition tested: two encoding schemes, four connectivity densities, five excitability regimes, and learning rates spanning zero to five hundred times the default. A decoder-independent measure of category separability shows STDP's net contribution to be approximately zero and, on average, slightly negative; separability is present at initialization and decreases during training. The framework's above-chance performance (~68%) is produced entirely by the Hebbian readout layer reading a fixed, pre-existing separability the organoid possesses before any learning occurs. The root cause is identified: under rank-order encoding each electrode emits a single spike, and the LIF population sits at roughly three times firing threshold with negligible inter-neuron spread, so every neuron fires exactly once regardless of synaptic weight. The population code collapses to a scalar spike count rather than a discriminable spatial pattern, leaving STDP nothing to shape. We further show that the freeze-weights test proposed in v1.0 as an attribution standard does not perform that function: it holds the decoder fixed alongside the organoid weights and therefore cannot separate their contributions. It is also inapplicable to living tissue, since biological synapses cannot be frozen. Replacement controls that do transfer to hardware are proposed. We report this as a negative result. The framework is sound as an engineering artifact and its mechanisms are correctly implemented; what it does not do is demonstrate STDP-driven categorical learning. We argue that this failure mode — apparent closed-loop learning entirely attributable to the readout layer — is a general hazard in organoid intelligence work, and that the attribution controls used to exclude it require more scrutiny than they have generally received.