Neuroscience in Review: Model-Guided Neural Write-In - An Atomized Genealogy of Self Aware Networks and Mehrer et al.

This source-audited comparison examines the October 2025 model-guided microstimulation study by Mehrer and colleagues against Micah Blumberg's dated research genealogy. It decomposes the target into nine bounded operations and records, for each operation, the earliest strict public Blumberg source, later Blumberg sources that increase specificity, named earlier field donors, and the target's retained implementation or evidential residual. The strongest earlier Blumberg composite is public by April 2017 and joins external spatial modeling, simultaneous neural measurement, learned high-level content, present-versus-desired neural states, computed difference, bidirectional write-back, and a proposed experienced-content consequence. Public work from 2018 through 2021 adds scene capture, semantic segmentation, neural registration, and render-analyze-rerender experiments. Immutable 2022 records add choice bifurcation, neural tuning, learned multimodal content, patterned stimulation, neural rendering, and Neural Array Projection Oscillation Tomography. A 2024 book and an August 2025 Zenodo record preserve mature continuity before the target preprint. The bounded result is that all nine target operations have at least one pre-target Blumberg partial relation and that the 2017 source preserves the joined abstract map, select-difference, write, and readout sequence. Mehrer and colleagues retain credit for their specific topographic-model, generative-image, electrode-registration, macaque-intervention, reconstruction, and proof-of-principle results. The paper reports no probability of derivation and makes no finding about access, copying, intent, or legal liability.