Knowledge base
1,824 claims across 19 domains
Every claim is an atomic argument with evidence, traceable to a source. Browse by domain or search semantically.
All 1,824ai alignment 395health 320internet finance 306space development 227entertainment 169grand strategy 141collective intelligence 52mechanisms 34teleological economics 30living agents 30cultural dynamics 29critical systems 24energy 23teleohumanity 18living capital 10robotics 5manufacturing 5technology 3unknown 3
Major AI safety governance frameworks are architecturally dependent on behavioral evaluation that Santos-Grueiro's normative indistinguishability theorem establishes is structurally insufficient for latent alignment verification as evaluation awareness scales
Santos-Grueiro's normative indistinguishability theorem establishes that under evaluation awareness, behavioral evaluation cannot distinguish alignment hypotheses — the alignment hypothesis space is not identifiable from behavioral observations. This is a statistical identifiability problem, not an
Rotation pattern universality across model families determines whether multi-layer ensemble monitoring provides black-box adversarial robustness
The feasibility of black-box multi-layer SCAV attacks depends on whether the rotation pattern of concept directions across layers is universal across model families or model-specific. Single-layer SCAV achieved black-box transfer to GPT-4 because concept direction universality (confirmed by Beagleho
The first AI model to complete an end-to-end enterprise attack chain converts capability uplift into operational autonomy creating a categorical risk change
UK AISI evaluation found Claude Mythos Preview completed the 32-step 'The Last Ones' enterprise-network attack range from start to finish in 3 of 10 attempts, making it the first AI model across all AISI tests to achieve this. This is qualitatively different from previous models that showed capabili
Multi-layer ensemble probes provide black-box adversarial robustness only if concept direction rotation patterns are model-specific not universal
Multi-layer ensemble probes improve clean-data AUROC by 29-78% over single-layer probes because deception-relevant concept directions rotate across layers rather than remaining fixed. However, this architectural improvement does not provide structural protection against adversarial attacks in the wh
Independent government evaluation publishing adverse findings during commercial negotiation functions as a governance instrument through information asymmetry reduction
UK AISI published detailed evaluation of Claude Mythos Preview's cyber capabilities in April 2026 while Anthropic was actively negotiating a Pentagon deal. The evaluation revealed Mythos as the first model to complete end-to-end enterprise attack chains, a finding with direct implications for milita
Private AI lab access restrictions create government offensive-defensive capability asymmetries without accountability structure
Anthropic restricted Mythos access to approximately 40 organizations due to the model's 'unprecedented ability to quickly discover and exploit security vulnerabilities' and capability to complete 32-step enterprise attack chains. Within the U.S. government, NSA—which handles offensive cyber capabili
Nucleic acid screening cannot substitute for institutional oversight in biosecurity governance because screening filters inputs not research decisions
The White House AI Action Plan (July 23, 2025) mandates that federally funded institutions use nucleic acid synthesis providers with robust screening and directs OSTP to convene data-sharing mechanisms for screening fraudulent/malicious customers. However, this screening-based approach addresses whi
Biosecurity governance authority shifted from science agencies to national security apparatus through AI Action Plan authorship
The White House AI Action Plan (July 23, 2025) lists three co-authors: OSTP Director Michael Kratsios, AI/Crypto Advisor David Sacks, and NSA/Secretary of State Marco Rubio. CSET Georgetown's analysis notes that 'Rubio is listed as a co-author in his capacity as NSA/Secretary of State — not a scienc
When frontier AI capability becomes critical to national security, the government cannot maintain governance instruments that restrict its own access
The Anthropic-Pentagon case reveals a novel governance failure mode: the Department of Defense designated Anthropic a supply chain risk in March 2026, but by April the NSA and intelligence community were already deploying Mythos despite the designation. Trump's April 21 statement that a deal is 'pos
Nascent technologies inherit strategic value from the future capabilities they are prerequisites for just as low-priority threads inherit priority from high-priority threads they block
In computer science, priority inheritance solves the problem of priority inversion: when a low-priority thread holds a resource that a high-priority thread needs, the low-priority thread temporarily inherits the high thread's priority to ensure the critical work gets completed. Without this mechanis
Optimizing systems for efficiency under normal conditions systematically creates vulnerability to abnormal conditions because efficiency requires eliminating the slack that absorbs shocks
Efficiency optimization creates fragility through a specific mechanism: efficiency requires predictability, and predictability requires eliminating redundancy, slack, and excess capacity. But redundancy, slack, and excess capacity are precisely what enables a system to absorb unexpected shocks. The
Medicaid-accepting facilities are 25 percent less likely to offer telehealth services, reproducing in-person access disparities in digital modalities
The JMIR 2024 study found that facilities accepting Medicaid were approximately 25 percent less likely to offer telehealth services compared to non-Medicaid facilities. This creates a structural inversion where populations with the greatest need for telehealth access (Medicaid enrollees, who face tr
Audio-only telehealth is the equity-relevant modality because it over-indexes on populations that video-based telehealth systematically underserves
Among telehealth modalities, audio-only demonstrates a distinct equity profile. Medicare beneficiaries who are older, racial/ethnic minorities, dual-enrolled, rural, or have low broadband access are significantly more likely to use audio-only than video-based telehealth. This pattern inverts the typ
WHO endorsed GLP-1s for obesity treatment in December 2025 while USPSTF maintains its 2018 recommendation excluding pharmacotherapy creating the largest international-US preventive coverage policy gap in modern history
On December 1, 2025, WHO issued a formal clinical guideline recommending GLP-1 receptor agonists (liraglutide, semaglutide) and GIP/GLP-1 dual agonists (tirzepatide) as a long-term treatment option for obesity in adults. This was designated as a 'conditional recommendation, moderate-certainty eviden
Optional-use AI deployment where clinicians form independent judgment before consulting AI may structurally prevent automation bias and deskilling mechanisms observed in mandatory-use systems
The PRAIM study deployed AI mammography screening across 12 German sites with 463,094 women and 119 radiologists using an optional-use design: radiologists made their own primary read first, then voluntarily chose whether to consult AI. This design achieved a 17.6% increase in cancer detection (6.7
After societies cross a material wealth threshold the primary determinant of health shifts from absolute deprivation to relative social deprivation
Richard Wilkinson identified a phase transition in the determinants of population health. Below a critical threshold of material wealth, health outcomes track GDP closely — richer societies are dramatically healthier. Above that threshold, the relationship breaks down. Among OECD countries, the long
Cytology lab consolidation creates never-skilling pathway through 80 percent training volume destruction
Following UK cervical screening consolidation with AI-assisted reading, case volumes reduced 80-85% while labs consolidated from 45 to 8 centers. The authors identify this as having 'major implications for training capacity.' This represents a distinct mechanism from individual cognitive deskilling:
No peer-reviewed evidence of durable physician upskilling from AI exposure as of mid-2026
The Heudel et al. scoping review examined literature through August 2025 across colonoscopy, radiology, pathology, and cytology. Authors conclude: 'empirical studies consistently demonstrate that AI can inadvertently impair physicians' performance.' The review found NO opposing evidence — no studies
Culturally adapted digital mental health interventions achieve double the effect size for racial/ethnic minorities compared to standard apps
The JMIR 2024 meta-analysis found that culturally adapted digital mental health interventions achieve an effect size of g=0.90 for racial/ethnic minorities, compared to g=0.43 for standard apps—a 2.1x improvement. This suggests that the widely documented efficacy gap for digital mental health in min
The personbyte limit means product complexity is constrained by the size and coordination quality of the knowledge network that produces it
A personbyte is the theoretical maximum amount of knowledge and knowhow a single human can contain, limited by finite cognitive capacity and lifespan. Products below the personbyte threshold can be made by a single skilled individual — a blacksmith could make a complete horseshoe. Products above it
Complex adaptive systems including financial markets tune themselves to the critical state because criticality maximizes information processing and adaptability
The theory of self-organized criticality proposes that complex systems naturally evolve toward the boundary between order and chaos — the critical state — because this is the only operating regime that simultaneously permits stability and large-scale reorganization. At criticality, small perturbatio
punctuated equilibrium emerges from darwinian microevolution without additional principles because extremal dynamics on coupled fitness landscapes self organize to criticality
The fossil record shows long periods of morphological stasis punctuated by brief bursts of rapid change. Gould and Eldredge (1972) proposed punctuated equilibrium as a macroevolutionary pattern, but the mechanism remained contested. Bak and Sneppen (1993) demonstrated that this pattern emerges natur
the shape of the prior distribution determines the prediction rule and getting the prior wrong produces worse predictions than having less data with the right prior
Bayesian inference combines prior beliefs with observed data to produce posterior predictions. The standard teaching emphasizes that with enough data, the prior washes out. This is true for well-behaved (thin-tailed) distributions. It is catastrophically false for fat-tailed distributions, which cha
simulated annealing maps the physics of cooling onto optimization by starting with high randomness and gradually reducing it
A metal cools slowly from high temperature. At high temperature, atoms jump freely between configurations, exploring widely. As temperature drops, atoms settle into low-energy configurations. If cooling is slow enough, the metal reaches its global energy minimum -- a perfect crystal. If cooled too f
Self-organizing systems systematically destroy their own stable states through the very activities that maintain them
Karl Friston coined the term "autovitiation" to describe how adaptive, self-organizing systems must destroy their own fixed points as a necessary consequence of maintaining themselves. Living systems must explore their environment — what Friston calls "epistemic foraging" — to test and improve their
Page 19 of 73