Intelligence Is Structure, Not Scale: A Whole Central Nervous System Connectome as a Fixed Substrate for Scoring Health Data Trust

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

We took the wiring diagram of a fruit fly's central nervous system, held every connection and sign fixed, and trained only synaptic gains, an input projection, and a readout. The task: reproduce two SuperTruth trust scores, the Data Trust Index (DTI) on health records and the Behavioral Integrity Index (BII) on agent event logs. Controls: a shuffled graph, a random graph, a parameter-matched network, a linear readout, and four frontier models (Claude Opus 5, GPT-5, Grok 4, Gemini 3 Flash) given the published DTI paper and the same records. On the first of five seeds, the degree-preserving shuffle matched the fly's wiring against the DTI engine (composite error 1.53 against 1.58 points); the random graph matched on tier agreement and every dimension. Every fixed graph beat the parameter-matched network by 1.9 to 2.2 points: provisional until five seeds; the substrate carries the computation. On 300 identical records, the four models matched the engine's tier 20% to 45% of the time, in 13 s to 73 s and at .04 to .26 cents per record; the fly, 84% at 16 ms and with no marginal cost. To our knowledge, as of 20 September 2026, this is the first reported use of a whole central nervous system connectome to score the trustworthiness of health data. Every record was synthetic; we used no real person's data. Simply put, the future of intelligence is analog. Version 1.0 is provisional: it reports seed 1 of the pre-registered five; later versions add the remaining seeds, the BII controls, and a post-hoc examples arm at the same DOI.