Notes from the Field

Notes from the Field

by Alexander Stewart
Season 1

The Weird Things That Correlate With Intelligence

AI
Higher childhood cognitive scores have been linked to greater odds of later reporting use of some illegal drugs. Higher early-life scores have also predicted lower mortality. Why can both findings be true—and what would be wrong with combining them into a personality type? Starting with the height–IQ question, this conversation explores brain volume, nearsightedness, curiosity, drug experimentation, longevity, vegetarianism, trust and an uncertain semen-quality finding. Small effects, imperfect labels and replication change the story. This is a public science discussion, not a way to estimate anyone’s IQ. Associations do not establish causes. Sources were independently checked; synthetic dialogue was generated with NotebookLM and edited and reviewed for accuracy. Sources Height: Beauchamp et al. (2011) Brain volume: Pietschnig et al. (2022) Myopia: Megreli et al. (2020) Personality: Anglim et al. (2022) Drug use: White, Gale & Batty (2012) Separate drug cohort: White & Batty (2012) Mortality review: Calvin et al. (2011) Vegetarianism: Gale et al. (2007) Trust: Sturgis, Read & Allum (2010) US vocabulary/trust: Carl & Billari (2014) Semen: Arden et al. (2009) Semen replication: DeLecce et al. (2020)

Gold, Empire, and the Price of Civilization

AI
What did a year of Roman military service cost in gold—and what does that reveal about the economic foundations of power? Starting with a legionary's roughly 2.25 troy ounces of gold-equivalent basic annual cash pay, this Notes from the Field episode follows a larger question: how do capital, energy, productive surplus and fiscal capacity change the cost of civilization? We explore: • Why the modern soldier is a node in a vast capital and logistics system. • Why gold cancels out of defense spending divided by GDP. • Why identical military/GDP ratios can hide very different economic strains. • Six lenses on imperial strain: output, surplus, revenue, interest, usable reserves and geopolitical effectiveness. • Gold purchasing-power yardsticks for labor, food, energy and land—and what AI might change. This is a research framework, not a completed two-thousand-year data set. Constant gold wages and a universal imperial-collapse signature remain hypotheses to test. Ancient estimates are uncertain; basic pay is not total compensation; reserve coverage is not literal wartime endurance. Selected sources: Roman economic and military-pay estimates: https://www.roiw.org/1984/263.pdf BEA defense/GDP series: https://fred.stlouisfed.org/series/A824RE1A156NBEA Historical military spending: https://eh.net/encyclopedia/military-spending-patterns-in-history/ Gold purchasing-power caution: https://www.nber.org/papers/w18706 Ideas and source narrative: Alexander Stewart. Dialogue generated with NotebookLM using synthetic hosts and edited for length and accuracy; this is not a recording of Alexander speaking. General historical and economic commentary, not individualized investment advice or a recommendation to buy gold.

A Thousand AI Agents in Runescape

AI
Max Bittker’s RS-SDK demo contains no LLM inference and is not Alex’s experiment. It is not evidence of “AI society”: the unit appears to be operator-plus-fleet, not independent agents. Includes MIT-licensed LostCity. Dialogue: NotebookLM.

The Middle Class Was a Fluke

AI
Was the postwar middle class a natural endpoint of development—or a contingent bargaining equilibrium? This episode asks whether broad prosperity expanded partly because states and firms depended heavily on mass participation as soldiers, workers, and consumers, and what happens when that dependence weakens. Its strongest contribution is a testable bargaining-power question: who needs whom, for what, and how do technology, globalization, professional militaries, mobile capital, and political institutions change that leverage? The episode compresses a large and contested history. It does not establish war as the primary cause of equality, coordinated elite intent, or a deterministic future under automation. Those claims require comparative historical research and counterexamples. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

The Invariants: A Strategy for Stability in 2036

AI
Forecasts about 2036 will mostly be wrong. This episode tries a different method: build strategy around forces that may persist even as technology changes—status, incentives, accountability, compounding, institutional delay, scarce human time, and emotional adaptation. Its strongest question is what becomes scarce when machine intelligence becomes abundant. The proposed answer is responsibility: the human or institution willing to sign, bear losses, and answer for consequences. The “invariants” are working hypotheses, not universal laws; the value is in testing where each one holds and where it breaks. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

The Stimulant Scarcity: Navigating the ADHD Medication Crisis

AI
Why can patients fail to obtain a medicine even when a national manufacturing ceiling appears to leave room? This episode maps interacting constraints across manufacturers, ingredients, distributors, pharmacies, insurers, regulation, and a low-margin generic-drug market that may underinvest in redundancy. The durable systems insight is that raising one legal ceiling cannot by itself repair physical supply, distribution throttles, formulation-specific shortages, or weak incentives for resilience. Safety note: this episode contains time-sensitive medical, legal, pharmacy, and insurance claims and should not be used to make medication decisions. Shortage status and dispensing or transfer rules can change by drug, dose, jurisdiction, pharmacy, and insurer. Use a licensed clinician or pharmacist and current official sources. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

Peak People: China’s Demographic Cliff and Strategy of Decline

AI
How does a large, aging, and shrinking country substitute capital and automation for people? This episode examines China's durable demographic constraint and the possible shift from labor quantity toward robotics, capital intensity, later retirement, and higher output per worker. The useful question is not whether demography mechanically determines China's future. It is which institutions and technologies can change productivity, participation, and dependency—and how quickly they can diffuse. Population and fertility figures are date-sensitive, estimates differ, and the episode's geopolitical interpretations are hypotheses rather than consequences implied by demography. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

When Plumbers Outearn Knowledge Workers

AI
Which forms of human value become scarce when routine cognitive output becomes cheap? This episode explores skilled physical work, verified origin, long track records, energy infrastructure, judgment, and accountability as possible sources of relative value in an AI-rich economy. Its strongest idea is the provenance premium: when polished synthetic output is abundant, a durable record of who produced something, how it was produced, and whether the source has been right before may matter more than polish itself. The title is provocative, not a general national wage fact: broad U.S. data do not show plumbers outearning software developers or computer occupations overall. Occupational, energy, and timing claims require specific geography and current evidence. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

The Horizon of 2041: The World Your Child Inherits

AI
How should someone prepare for 2041 when point forecasts about AI, housing, energy, demography, and institutions are likely to fail? This episode uses scenarios to separate volatile assumptions from more durable needs: judgment, relationships, meaning, adaptability, and the ability to become useful. Its strongest unresolved question is the apprenticeship bottleneck. If organizations automate junior cognitive work but still need experienced human judgment, how does a novice acquire the experience required to become senior? The scenario probabilities and many numerical projections are illustrative, not outputs from a reproducible forecasting model. They should not be relied upon as predictions. Sources were curated by Alex; dialogue audio was generated with NotebookLM.

The Architect of Autonomy: Chasing the First AI Unicorn

AI
If AI systems are already highly capable, why has no visibly autonomous agent built a billion-dollar company? This episode argues that the binding constraints may be institutional rather than cognitive: legal accountability, banking and identity requirements, trust, operational reliability, and the ability to recover from long-tail errors. It distinguishes three ideas that are often blurred together: an AI-led company, an AI-operated company, and an AI-owned company. The likely first “AI unicorn” may look conventional from the outside—a small human legal shell—while autonomous systems perform much of the work underneath. Timelines in the discussion are hypotheses, not forecasts. Sources were curated by Alex; dialogue audio was generated with NotebookLM.
1 of 2