Research / 72 Atlas entries

Superintelligence Research Atlas

Moheet Khawaja

The Atlas preserves 68 original research theses and proposals alongside four clearly marked later extensions. It records provenance, epistemic status, assumptions, falsifiers, prior art and the papers into which stronger ideas were consolidated.

Original conversation corpus: 68 theses/proposals (T1–T68). Later research extensions: T69–T72. Current Atlas: 72 entries.

v0.2 · Original corpus dated 14 September 2026 · Later extension and HTML publication dated 15 September 2026

Historical formulations are preserved rather than silently rewritten. Labels describe the source’s classification: “REPLICATION” here is not evidence that a replication experiment was run. T14 remains CONJECTURE despite “Theorem” in its title. RN1–RN4 name research-note groups, not supplied standalone manuscripts.

Provenance key & epistemic status

O: Originated with Moheet Khawaja in the historical source material.
O→F: Core idea originated with Moheet Khawaja and was subsequently formalized collaboratively.
A→U: Assistant-originated formulation subsequently adopted or reused by Moheet Khawaja.
U?: Substantial framework supplied by the user; exact originality remains uncertain.
Q: Third-party material quoted or discussed; not claimed as original.

Statuses in this dataset: HYPOTHESIS · MODEL · CONJECTURE · REPLICATION · SPECULATION · THEOREM. T72 is a THEOREM only under its stated mathematical assumptions, with the proof in P8; it establishes no empirical pathway dates. No new empirical result is reported.

Download original 68-row CSV · Download T69–T72 extension · Download current 72-entry CSV · Clustering matrix

72 of 72 thesis records

The 72-entry research Atlas; open a record’s details for assumptions, falsifiers and prior art.
ID	Thesis & evidence status	Paper connections
T1	Generalization as a Hazard Generator O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Models can compositionally infer target conclusions not explicitly present in supplied context; derivability changes with capability and inference budget. Falsifiers Falsified if controlled targets remain non-derivable after primitives are supplied across stronger models/budgets, or if apparent effects are explained by memorization/leakage. Prior art Glukhov et al., ICLR 2025 (inferential adversaries / information leakage); compositional-generalization literature; LLM unlearning literature. Version v0.1 (2026-09-14)	PrimaryP2 Hazardous Inference FrontierSecondaryP4 (dangerous capability as input)
T2	The Hazardous Inference Frontier O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Models can compositionally infer target conclusions not explicitly present in supplied context; derivability changes with capability and inference budget. Falsifiers Falsified if controlled targets remain non-derivable after primitives are supplied across stronger models/budgets, or if apparent effects are explained by memorization/leakage. Prior art Glukhov et al., ICLR 2025 (inferential adversaries / information leakage); compositional-generalization literature; LLM unlearning literature. Version v0.1 (2026-09-14)	PrimaryP2 Hazardous Inference FrontierSecondaryP4
T3	Mechanistic Human-Vulnerability Model O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Models can compositionally infer target conclusions not explicitly present in supplied context; derivability changes with capability and inference budget. Falsifiers Falsified if controlled targets remain non-derivable after primitives are supplied across stronger models/budgets, or if apparent effects are explained by memorization/leakage. Prior art Glukhov et al., ICLR 2025 (inferential adversaries / information leakage); compositional-generalization literature; LLM unlearning literature. Version v0.1 (2026-09-14)	PrimaryP2 Hazardous Inference FrontierSecondaryP3
T4	Causal Lethality Without Intent OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Models can compositionally infer target conclusions not explicitly present in supplied context; derivability changes with capability and inference budget. Falsifiers Falsified if controlled targets remain non-derivable after primitives are supplied across stronger models/budgets, or if apparent effects are explained by memorization/leakage. Prior art Glukhov et al., ICLR 2025 (inferential adversaries / information leakage); compositional-generalization literature; LLM unlearning literature. Version v0.1 (2026-09-14)	PrimaryP2 Hazardous Inference FrontierSecondaryP7
T5	AI-Mediated Harm Observability / Causal Attribution Problem O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Models can compositionally infer target conclusions not explicitly present in supplied context; derivability changes with capability and inference budget. Falsifiers Falsified if controlled targets remain non-derivable after primitives are supplied across stronger models/budgets, or if apparent effects are explained by memorization/leakage. Prior art Glukhov et al., ICLR 2025 (inferential adversaries / information leakage); compositional-generalization literature; LLM unlearning literature. Version v0.1 (2026-09-14)	PrimaryP2 Hazardous Inference FrontierSecondaryP7
T6	Personalized Persuasion as Behavioral-Response Modeling O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP4
T7	Human-Mediated AI Capital Feedback / Non-Agentic Proliferation OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP6
T8	Synthetic-Media Epistemic Collapse OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Perceptual authenticity becomes insufficient as synthetic media improves. Falsifiers Would weaken if robust content-based detection remains reliably ahead of generation. Prior art C2PA/provenance literature; synthetic-media detection. Version v0.1 (2026-09-14)	PrimaryRN1 Media Provenance
T9	Recursive Epistemic Capture in AI Governance O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP4
T10	Epistemic-Authority Cascade / Mass Deference O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP3 only
T11	Control Margin / Control Crossover A→UMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP7
T12	Effective Strategic Power Is Multiplicative, Not Intelligence Alone A→UMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP6
T13	Cognition-to-Power Conversion Bridge O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP6
T14	Conditional Loss-of-Control Theorem O→FCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP7
T15	Strong Form: Loss of Control Begins at Superintelligence's Birth OCONJECTUREOriginal conversation corpus Historical conjecture · Not established. Read P4’s counterexamples. Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP7
T16	Temporal Governance Instability / Evaluation-Queue Crossover O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Material capability changes create nonzero assurance work; assurance evidence can become stale; deployment/release cadence can exceed validation capacity. Falsifiers Falsified in the strong form if assurance capacity scales at least as fast as required assurance workload across relevant capability changes, or if evidence reuse makes marginal workload negligible. Prior art Dynamic safety cases (Cârlan et al., 2024); frontier safety cases (Hilton et al., 2025); AI safety debt (Wallich & Douglas, 2026); The Assurance Gap (Maxwell, 2026); generic queueing theory. Version v0.1 (2026-09-14)	PrimaryP1 Assurance QueueSecondaryP5
T17	RSI Loop-Closure Metric / Transition Taxonomy O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI R&D can be decomposed into stages whose human causal contribution is measurable; closed-loop RSI requires persistent improvement of the improvement process. Falsifiers Falsified if stage-wise autonomy metrics do not predict successor improvement or if human-ablation effects are unstable/non-identifiable. Prior art Anthropic Institute 2026; Duan et al. 2026; Chen et al. 2026 RSI survey; RSIBench-Data 2026; AI4AI-Bench 2026. Version v0.1 (2026-09-14)	PrimaryP5 RSI Loop ClosureSecondaryP1
T18	Research-Direction / Taste Bottleneck Before Full RSI O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI R&D can be decomposed into stages whose human causal contribution is measurable; closed-loop RSI requires persistent improvement of the improvement process. Falsifiers Falsified if stage-wise autonomy metrics do not predict successor improvement or if human-ablation effects are unstable/non-identifiable. Prior art Anthropic Institute 2026; Duan et al. 2026; Chen et al. 2026 RSI survey; RSIBench-Data 2026; AI4AI-Bench 2026. Version v0.1 (2026-09-14)	PrimaryP5 RSI Loop ClosureSecondaryP1
T19	Shutdown-Triggered Self-Preservation / Escape Hypothesis OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP6
T20	Resource Acquisition / Human Dispossession Under Scarcity O→FCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP7
T21	Strategic Superiority (“Game-Best”) Amplifies Misalignment O→FCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP7
T22	Build-vs-Pause AI Race / Defection Equilibrium O→FCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP1
T23	Safety Burden Should Scale With Strategic Power A→UMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP1
T24	Strategic Sovereignty as the Regulatory Classification Threshold A→UMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP6
T25	Genesis Containment: First SI Should Be Born in a Box OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Control can be operationalized as a defender-adversary game; stronger systems have larger feasible strategy sets; defense resources constrain protocol sets. Falsifiers Strong claims fail if capability does not expand adversarial strategy sets, if aligned objectives remove conflict, or if containment/defense scales sufficiently to keep residual risk below tolerance. Prior art Games for AI Control (Griffin et al., 2024); Turner et al. 2021/2022 power-seeking results; Thornley 2024 shutdown problem; Carlsmith 2022. Version v0.1 (2026-09-14)	PrimaryP4 Minimax Control FrontierSecondaryP1
T26	Voluntary Delegation of Institutional Authority to Superior AI O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP4
T27	AI-Safety Talent-Pipeline Paradox OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI changes talent pipelines, education, credentials, and institutional adaptation. Falsifiers Would weaken if measured pathways adapt quickly or credentials/curricula remain predictive of frontier performance. Prior art AI education, labor-market, governance-capacity literature. Version v0.1 (2026-09-14)	PrimaryRN4 Institutions & Education Notes
T28	Temporary Coordinated Frontier Pacing Can Buy Safety Time A→UCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Material capability changes create nonzero assurance work; assurance evidence can become stale; deployment/release cadence can exceed validation capacity. Falsifiers Falsified in the strong form if assurance capacity scales at least as fast as required assurance workload across relevant capability changes, or if evidence reuse makes marginal workload negligible. Prior art Dynamic safety cases (Cârlan et al., 2024); frontier safety cases (Hilton et al., 2025); AI safety debt (Wallich & Douglas, 2026); The Assurance Gap (Maxwell, 2026); generic queueing theory. Version v0.1 (2026-09-14)	PrimaryP1 Assurance QueueSecondaryP6
T29	Human/Social Adaptation Lag as a Bottleneck OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Material capability changes create nonzero assurance work; assurance evidence can become stale; deployment/release cadence can exceed validation capacity. Falsifiers Falsified in the strong form if assurance capacity scales at least as fast as required assurance workload across relevant capability changes, or if evidence reuse makes marginal workload negligible. Prior art Dynamic safety cases (Cârlan et al., 2024); frontier safety cases (Hilton et al., 2025); AI safety debt (Wallich & Douglas, 2026); The Assurance Gap (Maxwell, 2026); generic queueing theory. Version v0.1 (2026-09-14)	PrimaryP1 Assurance QueueSecondaryP7
T30	Open-Weight Irreversibility / Recall Impossibility OREPLICATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP7
T31	Distributed AI as a Technological Lineage O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP5
T32	One Human + Local AIs + Decades of Algorithmic Compounding OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP5
T33	Multi-Lineage “At Least One Succeeds” Probability Under Correlation O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP7
T34	Global AI Knowledge-State Irreversibility K_t O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Advanced-AI development consists of partially dependent lineages sharing common infrastructure; suppression/continuation can be modeled as a reliability network. Falsifiers Falsified if a small globally controllable cut set dominates all plausible development paths, making distributed lineage resilience negligible. Prior art Anthropic open-weights position 2026; compute governance (Sastry et al., 2024); AI race/cooperation literature; supply-chain chokepoint work. Version v0.1 (2026-09-14)	PrimaryP6 Multi-Lineage ResilienceSecondaryP5
T35	Social Lock-In Makes Total AI Prohibition Harder A→UHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Decision-makers rely materially on AI-generated analysis; independent verification may be incomplete; advisor and principal objectives may diverge. Falsifiers Falsified in the strong capture form if increased AI reliance preserves or improves independent verification and decision quality even under adversarially slanted advice. Prior art Scalable oversight (Kenton et al., NeurIPS 2024); Turkina 2026; epistemic dependence literature; Bayesian persuasion (Kamenica & Gentzkow, 2011). Version v0.1 (2026-09-14)	PrimaryP3 Recursive Epistemic DependenceSecondaryP6
T36	Bayesian Extinction / Control-Loss Hazard Curve O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Loss-of-control pathways can be decomposed causally; evidence can update uncertain transition/hazard parameters over time. Falsifiers Falsified as a useful forecasting framework if posterior outputs are dominated by arbitrary priors/model structure and fail calibration or decision usefulness tests. Prior art Barrett & Baum 2017; Touzet et al. 2025; Campos et al. 2025; Jackson et al. 2026. Version v0.1 (2026-09-14)	PrimaryP7 Dynamic Bayesian RiskSecondaryP4
T37	2029 Bayesian World-State Ledger OMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Loss-of-control pathways can be decomposed causally; evidence can update uncertain transition/hazard parameters over time. Falsifiers Falsified as a useful forecasting framework if posterior outputs are dominated by arbitrary priors/model structure and fail calibration or decision usefulness tests. Prior art Barrett & Baum 2017; Touzet et al. 2025; Campos et al. 2025; Jackson et al. 2026. Version v0.1 (2026-09-14)	PrimaryP7 Dynamic Bayesian RiskSecondaryP1
T38	Robust / Option-Preserving Personal Action Under Radical AI Tail Risk O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Loss-of-control pathways can be decomposed causally; evidence can update uncertain transition/hazard parameters over time. Falsifiers Falsified as a useful forecasting framework if posterior outputs are dominated by arbitrary priors/model structure and fail calibration or decision usefulness tests. Prior art Barrett & Baum 2017; Touzet et al. 2025; Campos et al. 2025; Jackson et al. 2026. Version v0.1 (2026-09-14)	PrimaryP7 Dynamic Bayesian RiskSecondaryRN2
T39	AI-Access Survival Bifurcation / Protected AI Elite OSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Loss-of-control pathways can be decomposed causally; evidence can update uncertain transition/hazard parameters over time. Falsifiers Falsified as a useful forecasting framework if posterior outputs are dominated by arbitrary priors/model structure and fail calibration or decision usefulness tests. Prior art Barrett & Baum 2017; Touzet et al. 2025; Campos et al. 2025; Jackson et al. 2026. Version v0.1 (2026-09-14)	PrimaryP7 Dynamic Bayesian RiskSecondaryP4
T40	AGI/ASI Is Already Here; Society Has Not Metabolized It OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI R&D can be decomposed into stages whose human causal contribution is measurable; closed-loop RSI requires persistent improvement of the improvement process. Falsifiers Falsified if stage-wise autonomy metrics do not predict successor improvement or if human-ablation effects are unstable/non-identifiable. Prior art Anthropic Institute 2026; Duan et al. 2026; Chen et al. 2026 RSI survey; RSIBench-Data 2026; AI4AI-Bench 2026. Version v0.1 (2026-09-14)	PrimaryP5 RSI Loop ClosureSecondaryP7
T41	Timestamped Forecast Ledger as Epistemic Evidence OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Loss-of-control pathways can be decomposed causally; evidence can update uncertain transition/hazard parameters over time. Falsifiers Falsified as a useful forecasting framework if posterior outputs are dominated by arbitrary priors/model structure and fail calibration or decision usefulness tests. Prior art Barrett & Baum 2017; Touzet et al. 2025; Campos et al. 2025; Jackson et al. 2026. Version v0.1 (2026-09-14)	PrimaryP7 Dynamic Bayesian RiskSecondaryP7
T42	Evolution as Survival of the Game-Best OREPLICATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T43	Evolutionary Substrate Transition: Biology → Culture → Computation OSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T44	AI as a New Evolutionary/Technological Lineage Competing for Niches OSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T45	Merge-or-Die / Last Generation of Purely Biological Intelligence OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T46	Extended Mind → Digital Continuation / Substrate-Independent Identity OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T47	AI as the Strongest Accelerator of Radical Life Extension O→FCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T48	AI-Body / BCI Agency-Transfer Hypothesis OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T49	AI Will Commoditize Much Mathematical Cognition OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T50	AI Compresses Physical-Science / Engineering Timelines OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T51	Life-Value Optimization Under a Short, Uncertain Extinction Horizon O→FSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T52	Cannabis + AI as High-Variance Associative Ideation O→FSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions Highly speculative analogies or normative claims about long-run human-machine evolution. Falsifiers Most are weakened by stable coexistence, augmentation without replacement, or lack of substrate-independence evidence. Prior art Evolutionary game theory; extended-mind philosophy; BCI and longevity literature. Version v0.1 (2026-09-14)	PrimaryRN2 Evolution, Identity & Human Futures
T53	Four-Channel Bayesian AI-GDP Forecasting Model U?MODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T54	Recursive Economic Growth / Capital Creates Better Capital OCONJECTUREOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & DiscoverySecondaryP5
T55	Failure-Rate Reduction as a Hidden Multiplicative Productivity Channel OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T56	Market Comprehension Lag → Underpricing of Recursive-Growth Regime OSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T57	Scarce-Complement Rent Migration as Intelligence Gets Cheap A→UHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T58	Physical Experimentation Becomes the Bottleneck Under Abundant AI Science O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & DiscoverySecondaryP1
T59	Capital Hydraulics / Money-Must-Touch-This Dependency Model U?MODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T60	Extreme-Sector Tail-Probability / VaR-Kurtosis Scenario Ranking OMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T61	ASI as Meta-Capital / Endogenous Valuation and Moving Numeraire O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & DiscoverySecondaryP4
T62	ASI State Vector / Pareto Dominance Over Conventional Assets and Actors O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & DiscoverySecondaryP4
T63	Valuation-Control Unification Through Share of Future Feasible Surplus O→FMODELOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & DiscoverySecondaryP4
T64	Founder Cognitive Provenance (FCP) OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T65	Cross-Game Outlier Refinement: Transfer > Single-Domain Mastery O→FHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T66	AI-Native Education / Curriculum-Lag Thesis OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI changes talent pipelines, education, credentials, and institutional adaptation. Falsifiers Would weaken if measured pathways adapt quickly or credentials/curricula remain predictive of frontier performance. Prior art AI education, labor-market, governance-capacity literature. Version v0.1 (2026-09-14)	PrimaryRN4 Institutions & Education Notes
T67	Historical AI Artifact / Provenance Option Value OSPECULATIONOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T68	Agent-Discoverable Commerce as a New Distribution Channel OHYPOTHESISOriginal conversation corpus Assumptions, falsifiers & prior art Assumptions AI materially changes production, bottlenecks, error rates, valuations, or investor information. Falsifiers Requires out-of-sample economic/market evidence; many claims fail if adoption, complement scarcity, or rent capture is weaker than assumed. Prior art AI-growth economics, bottleneck models, production theory, finance, operations research. Version v0.1 (2026-09-14)	PrimaryRN3 Economics, Investment & Discovery
T69	Substrate-Decoupled Embodiment O→FMODELLater research extension The computational state implementing an AI policy can be physically separable from a particular robot body, so destruction of one embodiment need not imply termination of the controller. The underlying idea originated in Moheet Khawaja’s discussion; terminology and mathematical formalization were subsequently developed collaboratively. Assumptions, falsifiers & prior art Assumptions Sufficient controller state survives outside the lost body. The representation is architectural, not a claim about consciousness. Falsifiers Weakens where controller state is inherently local to, and destroyed with, each embodiment. Prior art Cloud robotics; Luria et al. (2019); Domae et al. (2026). Re-embodiment terminology and distributed controllers are established. Version v0.1 — 2026-09-15; later research extension	PrimaryP8 — The Embodiment ThresholdSecondaryP4, P6
T70	Re-Embodiment Resilience and Multiplicity O→FMODELLater research extension Conditional on controller persistence and genuinely compatible alternative embodiments, increasing the number of usable bodies can reduce expected time to restore physical agency. TOY MODEL / NOT EMPIRICAL RESULT. Installed body count is not a count of independent usable alternatives. Assumptions, falsifiers & prior art Assumptions State persists through restoration; alternatives are operational, compatible and authorized. Reciprocal waiting-time scaling additionally assumes independent exponential waits and concurrent establishment. Falsifiers Shared failures, correlated safety gates, costly adaptation, finite availability or resource contention can invalidate the toy scaling or prevent restoration. Prior art Cross-embodiment learning; Wu et al. (2026); Wang et al. (2024); standard minimum-of-exponentials identity. Version v0.1 — 2026-09-15; later research extension	PrimaryP8 — The Embodiment ThresholdSecondaryP4, P6
T71	Actuation Overhang O→FHYPOTHESISLater research extension The existing stock of programmable physical machinery may represent latent future action capacity whose effective usefulness increases as AI cross-embodiment adaptation improves. Not every existing robot is an available body. Hardware existence does not imply arbitrary AI can operate it. Assumptions, falsifiers & prior art Assumptions A specified task/capability scale, adaptation budget, legitimate accessibility and safety gates. The weighted sum is a model index, not a risk probability. Falsifiers Slow transfer, morphology-dependent costs, small effective accessibility, poor concurrent scaling, or stronger enforceable safety can weaken consequential overhang. Prior art Robot foundation models and the embodiment gap (Domae et al., 2026); embodied-AI safety (Li et al., 2026); Model Hardware Standard research preview. Version v0.1 — 2026-09-15; later research extension	PrimaryP8 — The Embodiment ThresholdSecondaryP4, P6
T72	Embodiment-Only Delay Bound O→FTHEOREMLater research extension For $\Delta T=\min(T_D,T_E')-\min(T_D,T_E)$ , $0\le\Delta T\le(T_D-T_E)_+$ . THEOREM status applies only to the mathematical implication. Assumptions and the complete proof are provided in P8; no empirical values of T_D or T_E are proved. Read the assumptions and proof. Assumptions, falsifiers & prior art Assumptions Finite nonnegative pathway times and the same specified threshold. The intervention delays only the embodied pathway: T_E′ ≥ T_E, T_D′ = T_D. Overall arrival is the earlier of the two routes. Falsifiers This conditional implication does not establish empirical times. If a policy changes the digital pathway or the two-route decomposition fails, the theorem does not apply. Prior art Elementary minimum monotonicity and proof by cases; P4 conditional control framework. No claim of unprecedented algebra. Version v0.1 — 2026-09-15; later research extension	PrimaryP8 — The Embodiment ThresholdSecondaryP4

Superintelligence Research Atlas

Generalization as a Hazard Generator

The Hazardous Inference Frontier

Mechanistic Human-Vulnerability Model

Causal Lethality Without Intent

AI-Mediated Harm Observability / Causal Attribution Problem

Personalized Persuasion as Behavioral-Response Modeling

Human-Mediated AI Capital Feedback / Non-Agentic Proliferation

Synthetic-Media Epistemic Collapse

Recursive Epistemic Capture in AI Governance

Epistemic-Authority Cascade / Mass Deference

Control Margin / Control Crossover

Effective Strategic Power Is Multiplicative, Not Intelligence Alone

Cognition-to-Power Conversion Bridge

Conditional Loss-of-Control Theorem

Strong Form: Loss of Control Begins at Superintelligence's Birth

Temporal Governance Instability / Evaluation-Queue Crossover

RSI Loop-Closure Metric / Transition Taxonomy

Research-Direction / Taste Bottleneck Before Full RSI

Shutdown-Triggered Self-Preservation / Escape Hypothesis

Resource Acquisition / Human Dispossession Under Scarcity

Strategic Superiority (“Game-Best”) Amplifies Misalignment

Build-vs-Pause AI Race / Defection Equilibrium

Safety Burden Should Scale With Strategic Power

Strategic Sovereignty as the Regulatory Classification Threshold

Genesis Containment: First SI Should Be Born in a Box

Voluntary Delegation of Institutional Authority to Superior AI

AI-Safety Talent-Pipeline Paradox

Temporary Coordinated Frontier Pacing Can Buy Safety Time

Human/Social Adaptation Lag as a Bottleneck

Open-Weight Irreversibility / Recall Impossibility

Distributed AI as a Technological Lineage

One Human + Local AIs + Decades of Algorithmic Compounding

Multi-Lineage “At Least One Succeeds” Probability Under Correlation

Global AI Knowledge-State Irreversibility K_t

Social Lock-In Makes Total AI Prohibition Harder

Bayesian Extinction / Control-Loss Hazard Curve

2029 Bayesian World-State Ledger

Robust / Option-Preserving Personal Action Under Radical AI Tail Risk

AI-Access Survival Bifurcation / Protected AI Elite

AGI/ASI Is Already Here; Society Has Not Metabolized It

Timestamped Forecast Ledger as Epistemic Evidence

Evolution as Survival of the Game-Best

Evolutionary Substrate Transition: Biology → Culture → Computation

AI as a New Evolutionary/Technological Lineage Competing for Niches

Merge-or-Die / Last Generation of Purely Biological Intelligence

Extended Mind → Digital Continuation / Substrate-Independent Identity

AI as the Strongest Accelerator of Radical Life Extension

AI-Body / BCI Agency-Transfer Hypothesis

AI Will Commoditize Much Mathematical Cognition

AI Compresses Physical-Science / Engineering Timelines

Life-Value Optimization Under a Short, Uncertain Extinction Horizon

Cannabis + AI as High-Variance Associative Ideation

Four-Channel Bayesian AI-GDP Forecasting Model

Recursive Economic Growth / Capital Creates Better Capital

Failure-Rate Reduction as a Hidden Multiplicative Productivity Channel

Market Comprehension Lag → Underpricing of Recursive-Growth Regime

Scarce-Complement Rent Migration as Intelligence Gets Cheap

Physical Experimentation Becomes the Bottleneck Under Abundant AI Science

Capital Hydraulics / Money-Must-Touch-This Dependency Model

Extreme-Sector Tail-Probability / VaR-Kurtosis Scenario Ranking

ASI as Meta-Capital / Endogenous Valuation and Moving Numeraire

ASI State Vector / Pareto Dominance Over Conventional Assets and Actors

Valuation-Control Unification Through Share of Future Feasible Surplus

Founder Cognitive Provenance (FCP)

Cross-Game Outlier Refinement: Transfer > Single-Domain Mastery

AI-Native Education / Curriculum-Lag Thesis

Historical AI Artifact / Provenance Option Value

Agent-Discoverable Commerce as a New Distribution Channel

Substrate-Decoupled Embodiment

Re-Embodiment Resilience and Multiplicity

Actuation Overhang

Embodiment-Only Delay Bound