Michael Darius Eastwood · Independent AI alignment researcher
Published
Some minds learn across many fields at unusual speed, sustain research projects over years, and produce output disproportionate to formal training, while at the same time struggling with routine administration, timekeeping, task switching, and low stimulation procedural work. Paper C names that pattern, and names the institutional error that treats visible capability as evidence that no support is needed.
Institutions often respond to capable neurodivergent people with a specific misjudgement. "You wrote a book. You built a research programme. Surely you do not need adjustments." The paper asks whether that inference is safe. It is not, and existing clinical guidance already says so. NICE CG142 directs professionals to record the discrepancy between intellectual functioning and adaptive functioning. Strengths based autism research treats the same neurodivergent trait as advantage or impairment depending on context.
The profile is described with six components: cross domain learning, monotropic hyperfocus, executive function asymmetry, self directed research intensity, institutional misrecognition, and the descriptive "polymathy" label. Three of the components are backed by strong existing literature (monotropism, twice exceptionality, autistic underemployment and burnout). One is weak by explicit disclosure: "polymathy" has no standing in psychiatric nosology and is used descriptively only. The load bearing claim is functional. Capability in one domain does not remove support needs in another. The Equality Act 2010 s.20 duty is triggered by functional disadvantage, not by the absence of achievement elsewhere.
This is a working paper, not a diagnostic category, and the paper is explicit about that. The framework is falsifiable by construction. If PNP cannot be operationally distinguished from ordinary giftedness plus autism plus ADHD, it collapses into them and must be retired. Whether it adds incremental predictive value above existing formulations has not yet been tested. The "polymathy" component carries the least evidential weight and is presented as such. The paper describes a single case study (its own author) as an illustrative subject, and single case illustrations do not establish population level claims.
Paper C shares a structural conclusion with the Hernandez Espinosa et al. paper in PNAS Nexus (April 2026, Volume 5 Issue 4, pgag076), which argues that neurodivergent cognition is a protective factor for AI alignment: uniform minds miss what different minds catch. Gumbau Mezquita extended that argument formally in June 2026 (arXiv:2606.28639) using Trakhtenbrot's Wall from finite model theory. Paper C is the descriptive companion to that emerging mathematical result. Neither of the external authors cites Eastwood; both converged independently on the structural point.
The paper is on OSF at DOI 10.17605/OSF.IO/6C5XB. The Hernandez Espinosa et al. paper is verifiable at PNAS Nexus 5(4) pgag076 (DOI 10.1093/pnasnexus/pgag076). Gumbau Mezquita's formal extension is arXiv:2606.28639. The Capability and Adjustment Fallacy is stated in one line and can be attacked directly. Any evidence that visible high performance is a reliable indicator of low support need would compromise the argument.
From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.