Meditations on Moloch

Author: Scott Alexander (Slate Star Codex) · Published: July 30, 2014 · URL: https://slatestarcodex.com/2014/07/30/meditations-on-moloch/

One-line: Why systems that everyone inside them hates keep going anyway: competition rewards whoever sacrifices a shared value for an edge, until every value that can be traded away has been. Alexander names that force Moloch and argues that only a coordinator stronger than every competitor can beat it.

Saved to Kyle’s Apple Note “Things To Internet” (c. 2026), as one of the canonical written ideas in the list Packy McCormick opens “welcome to not boring world” with (January 6, 2026): “Aggregation Theory was written. … I, Pencil. Meditations on Moloch. The Cathedral and the Bazaar. … All written.”

Summary

The frame: Ginsberg’s Moloch. The essay opens with the Moloch section of Allen Ginsberg’s Howl and asks why Ginsberg treats civilization as a single being. Alexander’s answer runs through a passage from the Principia Discordia. Malaclypse complains that “nobody wants” the world’s cruelties and “everybody hates” them, and the Goddess replies, “Oh. Well, then stop.” If everyone hates the current system, who keeps it going? Ginsberg’s answer is “Moloch.” Alexander says the image works “not because it’s correct” but because “thinking of the system as an agent throws into relief the degree to which the system isn’t an agent.” His model case is a made-up country where everyone must shock themselves eight hours a day, and anyone who refuses gets killed by everyone else. Every citizen hates it, and it lasts anyway, “for lack of a good coordination mechanism.”

The fourteen multipolar traps. Alexander says he’ll list “let’s say ten” real examples and ends up listing fourteen. In each one, the fix is obvious “from a god’s-eye-view” and impossible “from within the system”:

1. The Prisoner’s Dilemma, played by two players who keep landing on defect-defect.
2. Dollar auctions, where odd auction rules get someone to pay $10 for a dollar bill.
3. The fish farms, from his Non-Libertarian FAQ: a thousand farms share a lake, and every farm does better by turning off its pollution filter until the lake is ruined. He calls this “the core of my objection to libertarianism.”
4. The Malthusian trap: rats on an island breed up to carrying capacity, and the sects that give up art, restraint or foresight to survive take over.
5. Capitalism: under “sufficiently intense competition,” companies have to drop every value except profit or get beaten by rivals who did.
6. The Two-Income Trap: families bid against each other for houses in good school districts, and second jobs cancel each other out.
7. Agriculture: farming societies outcompeted hunter-gatherers whether or not farming made life better.
8. Arms races: no country can disarm on its own.
9. Cancer: a cell that grabs resources to copy itself flourishes until the body dies.
10. The “race to the bottom”: jurisdictions keep cutting taxes and rules to attract business.

Then comes a second kind of trap, where an outside force like social stigma holds competition back. These systems don’t collapse to subsistence. They just can’t improve:

11. Education: students, employers and colleges all chase prestige.
12. Science: no single scientist or journal gains by adopting better statistics or preregistration first.
13. Government corruption: corporate welfare survives because officials compete to get elected.
14. Congress: 9% approval overall, but most people approve of their own representative.

He puts the principle underneath all of them this way: “In some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before.” Our institutions aren’t a utopia because “it wasn’t designed by humans.” Incentives carve institutions the way gravity carves riverbeds. Section II ends in Las Vegas, where he writes: “Some people have mystical experiences and see God. There in Las Vegas, I saw Moloch.”

The four brakes. Section III takes a Discordia joke literally (“Time flows like a river. Which is to say, downhill”) and asks why we aren’t already back at subsistence. He gives “three bad reasons,” plus one good one:

• Excess resources. A whale carcass on the ocean floor briefly feeds everything around it. We live in “an age of whalefall,” Robin Hanson’s “dream time,” when there is enough slack for art, music, philosophy and love.
• Physical limitations. Starving slaves doesn’t make them work harder, and a woman can have only one baby per nine months.
• Utility maximization. Capitalism and democracy do optimize for human values, but only while profit stays tied to satisfying customers and workers, and elections stay tied to good policy. Alexander calls that link fragile. It snaps as soon as a hidden pesticide, a labor glut or a coffee-picking robot uncouples the two.
• Coordination. “The opposite of a trap is a garden.” Governments, social codes, guilds, religions and corporations all bundle agreements with enforcement. But those coordinators face their own traps, which leaves a trade-off “between discoordination and tyranny.” A monarch is “an unincentivized incentivizer.” The price is being ruled by one man’s whims.

Technology takes the brakes away. Section IV argues that technology is “all about creating new opportunities,” including new opportunities to sell out a value. It removes physical limits (Soylent, modafinil, robots). It eats up spare resources once agents can be copied. It cuts capitalism loose from human values once workers stop being needed. And it makes propaganda more efficient (“technology increases the efficiency of manufacturing consent”). Coordination is the one brake technology could strengthen, but superintelligence would by definition hold more than 51% of the power. Left alone, “the river reaches the sea” in one of two places. One is Eliezer Yudkowsky’s nightmare of a superintelligence optimizing for something arbitrary. The other is Robin Hanson’s competition among copyable emulated minds, which could strip away even the wish for human values. In either case the last thing to go is consciousness itself, “having the lights on inside.”

Gnon, the Gods of the Copybook Headings, and the walled garden. Sections V and VI take on the neoreactionary answers. Nick Land wants us to comply with “Gnon” (Nature and Nature’s God). Alexander renames Kipling’s Gods of the Copybook Headings “GotCHa”: do what they demand, and you still die. Warg Franklin makes the strongest case for authoritarianism, a planned “walled garden” that captures Gnon. Alexander says he agrees with Franklin “more than I have ever agreed with anyone else about anything,” and then rejects the conclusion. Any garden still has to compete with everyone outside its walls.

The Gardener and Elua. Section VII is his answer: “The only way to avoid having all human values gradually ground down by optimization-competition is to install a Gardener over the entire universe who optimizes for human values.” Following Bostrom’s Superintelligence, he thinks a machine smarter than us that shares human values could do that job. “I am a transhumanist because I do not have enough hubris not to try to kill God.” Section VIII sets Moloch against Elua, the god from the Kushiel novels “of flowers and free love and all soft and fragile things. Of art and science and philosophy and love.” We should placate the other gods only as much as it takes “to avoid starvation and invasion,” and worship Elua. The essay closes on Moloch’s offer (“throw what you love most into the flames, and I can grant you power”) and on the line “Only another god can kill Moloch.”

Full text

Archived privately against link rot: ../attachments/meditations-on-moloch/meditations-on-moloch.md, along with the essay’s two images.

Connections

• Moloch: the concept page for the multipolar trap and race to the bottom, built from this essay alone.
• Game Theory and Prisoner’s Dilemma: the Prisoner’s Dilemma and the dollar auction are traps 1 and 2, and “god’s-eye-view versus within the system” is the essay’s version of the gap between individual and collective rationality.
• Tragedy of the Commons: the fish-farm lake (trap 3) is a commons problem, and Alexander uses it as his main case against leaving coordination to voluntary pacts.
• Capitalism: capitalism is trap 5, and it is also one of the four brakes. It serves human values only while profit stays tied to satisfying people. Alexander insists this is a different claim from “capitalists are greedy.”
• Existential Risk and AI Alignment: the essay’s endpoint is a superintelligence optimizing for the wrong thing, and its hope is one that shares human values.
• Nick Bostrom and Robin Hanson: Superintelligence supplies the Gardener argument, and Hanson supplies both “this is the dream time” and the em-competition nightmare.
• The Education of the Broligarchy: a tension worth keeping in view. Blake Smith’s Colossus essay on the Silicon Valley canon estimates that the canon can be read in a year, “provided, that is, you follow a version of the canon that excludes the prolific collected blog posts of thinkers like Scott Alexander and Eliezer Yudkowsky.” Smith’s canon leaves this blog out, while Packy McCormick’s January 2026 list of canonical written ideas puts it next to Aggregation Theory (07.21.15) and The Bitter Lesson. The disagreement is over what a canon is. Smith means books, and McCormick means ideas that got written down.
• The Bitter Lesson and Aggregation Theory (07.21.15): its neighbors on that list.