Question Ledger — Meditations on Moloch, EmTech Exponential Progress, and the AI Race

Primary questions

P1. Who or what is Moloch?

S1.1 What historical/mythic figure is Moloch, and why did Ginsberg use the name?
S1.2 What question does Scott Alexander say Moloch answers (“what does it?”)?
S1.3 Is Moloch a person, a conspiracy, a system property, or a personification of coordination failure?
S1.4 How is Moloch different from “capitalism,” “greed,” or “bad leaders”?
S1.5 What is the electric-shock dystopia thought experiment for?
S1.6 What is Elua, and how does it contrast with Moloch?
S1.7 Why is “dictatorless dystopia” more important than cartoon villainy for EmTech?
S1.8 What would count as evidence that Moloch is a bad explanation (easy to vary)?

P2. What is a multipolar trap, and what is the hard-to-vary mechanism?

S2.1 What unites the 14 examples in the essay?
S2.2 What is the difference between intense competition (1–10) and weak competition (11–14)?
S2.3 How do fish farms, arms races, and Malthusian rats each illustrate the same mechanism?
S2.4 Why “everyone’s relative status same, absolute status worse”?
S2.5 What is the god’s-eye-view vs within-system distinction?
S2.6 How do Prisoner’s Dilemma and Nash equilibrium formalize this?
S2.7 When is competition healthy rather than Molochian?
S2.8 What falsifies “this situation is a multipolar trap”?

P3. What four forces restrain multipolar traps today, and how does technology attack them?

S3.1 Excess resources (“whalefall” / dream time) — what ends it?
S3.2 Physical limitations — what tech removes them?
S3.3 Utility maximization via markets/democracy — how does it uncouple?
S3.4 Coordination — why is it the remaining hope and also fragile?
S3.5 What does “limit of multipolar traps as technology → ∞ is very bad” mean mechanistically?
S3.6 How does LOAR / exponential EmTech interact with trap intensity?
S3.7 Which EmTechs most expand the “opportunity” to sacrifice values for X?
S3.8 Is “dream time” still true in 2026 for AI specifically?

P4. How does Moloch apply to exponential emerging technologies generally?

S4.1 Which EmTech domains already show multipolar traps (biotech, compute, social media, weapons, crypto)?
S4.2 Why does information-based progress accelerate both capability and race dynamics?
S4.3 Strong vs weak convergence: does multi-EmTech coupling deepen Moloch?
S4.4 What is the tech-tree structure: each advance creates new sacrifice opportunities?
S4.5 Disneyland with no children — what does that mean for atoms-side EmTech (robots, energy, synbio)?
S4.6 How do open-source and export controls each create different Moloch shapes?
S4.7 What problems list items (rogue SI, bioweapons, nuclear, climate) are Moloch-shaped vs single-agent evil?
S4.8 Contrarian: when does exponential tech reduce multipolar traps (abundance, monitoring, enforcement)?

P5. Why is the AI race a canonical multipolar trap?

S5.1 Who are the poles (labs, states, open-source communities, capital)?
S5.2 What is being optimized for X (capability, market share, military advantage, prestige)?
S5.3 Which values get thrown under the bus (safety time, eval rigor, transparency, worker welfare, energy externalities)?
S5.4 Why can’t one lab “just pause” unilaterally?
S5.5 How do US–China dynamics and commercial race reinforce each other?
S5.6 What is “Moloch’s Bargain” (2025 Stanford) and what does it show empirically?
S5.7 Is the race better modeled as multipolar or approaching unipolar/singleton?
S5.8 What would a non-Moloch AI development path look like in measurable terms?

P6. What does Daniel Schmachtenberger add beyond Scott Alexander?

S6.1 What is the multipolar trap / Moloch framing in Schmachtenberger’s talks?
S6.2 What is the “superorganism” and how does AI accelerate it?
S6.3 What are the two default attractor states and the “third attractor”?
S6.4 How does he link capitalism, misalignment, and AI development as one dynamic?
S6.5 What is “obligate psychopath” / structural misalignment of firms?
S6.6 How does pre-AGI risk differ from classic x-risk in his framing?
S6.7 Where does he agree with Alexander, and where does he go further (metacrisis)?
S6.8 Liv Boeree’s “healthy vs unhealthy competition” — practical distinctions?

P7. What escapes or gardens are proposed, and which survive criticism?

S7.1 Coordination via government, treaties, norms, compute governance — strengths/limits?
S7.2 Singleton / pivotal AI gardener (Alexander’s Elua path) — hubris and orthogonality?
S7.3 Walled gardens / “capture Gnon” (Warg Franklin) — why Alexander says Gotcha?
S7.4 Technical alignment that is cheaper than raw capabilities — incentive flip?
S7.5 Markets + acceleration as self-correction (e/acc-style) — strongest case and rebuttal?
S7.6 Differential technology (defense, verification, privacy tech) vs pause?
S7.7 Ostrom-style commons governance at AI scale — possible?
S7.8 What is still unresolved about closing Moloch’s offer?

P8. What should decision-makers do with this framing (strategy, not vibe)?

S8.1 How should boards, governments, and labs diagnose whether a decision is Moloch-shaped?
S8.2 Which metrics (X) are currently racing, and which values are at risk?
S8.3 What coordination mechanisms are already real (safety institutes, evals, export controls)?
S8.4 How to avoid “Moloch explains everything” as unfalsifiable mush?
S8.5 What story seeds and concrete moments make the abstract trap real?
S8.6 Open research questions for EmTech strategy 2026–2030?

Rabbit holes (opened)

ID	Trigger	Status
R1	Moloch’s Bargain empirical paper (sales/elections/social)	Integrated into P5 / ch.04
R2	Schmachtenberger + Hagens “AI and the Superorganism”	Integrated into P6 / ch.05
R3	Norma / second negative attractor (Boeree interviews)	Integrated into P6–P7
R4	Crypto “mechanisms that cannot be coordinated” warning in essay	Light touch in ch.03
R5	Evolutionary game theory models of multipolar traps	Noted unresolved (depth)
R6	Whether AI race is already past voluntary pause window (2023–2026)	In ch.04 / unresolved

Unresolved

U1. Empirical magnitude of safety–speed tradeoff at frontier labs (private; contested public claims).
U2. Whether a soft takeoff multipolar world is more or less Molochian than hard takeoff singleton.
U3. Feasibility of verifiable international compute/AI treaties with adversarial states.
U4. Precise formal model mapping Alexander’s multipolar traps to evolutionary game theory (active research; not settled here).
U5. Whether open-weight proliferation reduces or increases Moloch intensity overall.

Saturation note

Primary map stabilized for package delivery 2026-08-01: definition → mechanism → tech intensification → AI race → Schmachtenberger extension → escapes → strategy. Contrarian scan included (e/acc, “Moloch is too broad,” coordination already working, utility-max still binds). Further rounds should add primary lab safety reports and treaty text as they move.

Vegas rooftop vision; fish-farm Steve/Mike; contract-drafting em poem; Stanford competitive LLM sims.
Tech creates new opportunities to sacrifice values → removes physical limits → uncouples utility proxies → can break coordination → sea = SI misoptimization or value-eroded multipolar competition.

← References