The data in words
The Foundry's machinery mass grows from the seed by doubling every 60 d, a straight line on this log scale, then levels off at full scale.
| Milestone | Date | Mass |
|---|---|---|
| Seed lands on Mercury | January 2029 | 1,000 t |
| Can build the Veil | March 2032 | 5.4 × 10¹¹ kg |
| Can build the Ring | November 2032 | 1.0 × 10¹³ kg |
| Can build the Lattice | March 2033 | 5.1 × 10¹³ kg |
| Full scale | June 2033 | 1.3 × 10¹⁴ kg |
Principle
A factory doubles when it can make every one of its own parts, chips and robots included, from what is
around it. If each kilogram of machinery builds another kilogram in a fixed time T, the mass grows as
M(t) = M₀ · 2^(t/T). Nothing else about the plan scales like that, and nothing else needs to.
The doubling arithmetic. From a seed to the the plan needs is log₂ of their ratio, rounded up to a whole number: doublings, which give . At each, that takes . Two consequences follow from the shape of the curve:
- Most of the mass arrives at the end. Each doubling makes as much as all the doublings before it. In , a year after landing, the Foundry is only ; by it is ; and in the eighteen months after that it grows to . Almost everything the plan builds, it builds in the last year.
- Every day counts many times over. Full scale comes doubling periods after landing, so each day added to the doubling time moves full scale, the Lattice's full power and the first ice launch by days.
Power follows from the doubling time. Making a kilogram of machinery from raw regolith takes about
. Growing continuously, the machinery adds mass at
M · ln 2 / T, so each kilogram draws e · ln 2 / T: . At full
scale that is , all of it from the Foundry's own solar film.
Why Mercury. Sunlight there is at its mean distance from the Sun. Mercury's axis is almost upright (NASA), so crater rims near its poles see the Sun nearly all the time and escape the day side's heat. Its gravity is weak: escape speed is , low enough that mass drivers, not rockets, throw parts to orbit. And its silicate rock yields oxygen, silicon and metals by electrolysis.
Design
The seed is twelve machine types that, together, make all twelve. The loop closes on materials (rock in, every part out) and on energy (film made by the loop powers the loop).
The data in words
12 machine types form a closed loop. Each feeds the next, except that the mass drivers' parts leave for orbit; the Assemblers close the loop by building every type, so the loop copies itself. It doubles every 60 d, and 27 doublings take the 1,000 t seed to 1.3 × 10¹⁴ kg.
| Machine | Makes | Feeds |
|---|---|---|
| Diggers | Sorted regolith | Smelters |
| Smelters | Metals and oxygen | Every other line |
| Glass furnaces | Glass and optics | Film lines, Lithocells, laser emitters |
| Film lines | Solar film | Power for the loop; Petals and slats |
| Lithocells | Chips | Every machine's controller |
| Choir nodes | Compute and links | Coordination, the Mass Ledger |
| Coil lines | Motors | Everything that moves |
| Rolling mills | Sheet, beam and rail | Assemblers, mass drivers, Petal and slat frames |
| Mass drivers | Parts to orbit | The Lattice, the Veil, the Ring |
| Haulers | Site transport | The whole loop |
| Printers | Complex metal parts | Assemblers, bearings, housings |
| Assemblers | Every machine type | The whole loop |
| Machine | What it does | What it feeds |
|---|---|---|
| Diggers | Mine and sort regolith | Smelters |
| Smelters | Solar furnaces and molten-regolith electrolysis, splitting rock into oxygen, iron, silicon, aluminum, titanium and magnesium | Every other line |
| Glass furnaces | Glass, fiber and optics | Film lines, Lithocells, laser emitters |
| Film lines | Thin solar film at , the Foundry's own power and the stuff of the Lattice and the Veil | Power for the loop; Petals and slats |
| Lithocells | Chip fabs that write circuits directly with electron beams | Every machine's controller |
| Choir nodes | Compute and communications for the Choir control system | Coordination, the Mass Ledger |
| Coil lines | Aluminum wire and iron cores for induction and reluctance motors, with no rare earths | Everything that moves |
| Rolling mills | Sheet, beam and rail | Assemblers, mass drivers, Petal and slat frames |
| Mass drivers | Throw finished parts to orbit | The Lattice, the Veil, the Ring |
| Haulers | Move ore, parts and products across the site | The whole loop |
| Printers | Electron-beam additive manufacturing in vacuum, for complex metal parts | Assemblers, bearings, housings |
| Assemblers | Robot arms that build every machine type, including Assemblers | The whole loop |
Simple mass, fast copies. A loop doubles only as fast as its slowest line. Ouroboros keeps the slow lines small: about of each copy's mass is complex parts (chips, bearings, optics); the rest is film, beams, rail and wire, which are quick to make in bulk. The seed carries the few trace elements it cannot yet refine, until refining closes that last gap in 2030 (the specifications below give the closure).
Products are diverted growth. The Foundry can build a product in one doubling period once its own mass equals the product's, because one doubling's output is the Foundry's mass. That sets when each piece of the plan becomes affordable:
- The Hesperus Veil, , on .
- The Braking Ring, , on .
- The Helios Lattice, , on .
Building the Veil at its close date, , costs the Foundry of growth: by then the Veil is a quarter of one doubling's output. Version 2's lighter Veil, built earlier, cost about a whole doubling. Matter is cheap once the loop runs: all of the plan's hardware takes about , less than a millionth of the energy the plan spends on water and cooling.
Specifications
| Parameter | Value |
|---|---|
| Seed mass | , launched , landed |
| Doubling time | |
| Doublings to full scale | , giving |
| Full scale | |
| Energy per kilogram built | |
| Power per kilogram of machinery | |
| Peak power | , from its own film |
| Share of Mercury used | of its mass |
| Mercury's escape speed | , so parts reach orbit by mass driver |
| Machinery in January 2030 | |
| Machinery in January 2032 | |
| Material closure | 99.99% at seed; 100% once trace-element refining runs in 2030 |
What is invented
Build sequence
- : Ouroboros-0 starts on Earth: a factory that builds itself from simulated Mercury regolith. It must show mass closure and a doubling time before the seed is committed. Under its kill signal, the plan falls back to the multi-century route.
- : The seed launches, one of the very few things in the plan that leaves Earth. The transfer to Mercury takes .
- : The seed lands at a polar site. The Choir goes live with it.
- : of machinery. Trace-element refining closes the loop completely.
- : The first Helios Petals fly and Veil slat production begins, small diversions from a growing base.
- : .
- : The Foundry can build the whole Veil in one doubling. It closes over Venus on .
- : The Braking Ring becomes affordable; its construction runs to .
- : The Helios Lattice becomes affordable.
- : Full scale, . The Lattice reaches full power the same day and the first Ice Bricks leave Ceres ().
Failure modes and safeguards
- Runaway replication. A machine that copies itself must stop when told. Every build order is signed by the Choir, and machines refuse unsigned orders in hardware. The Choir's Mass Ledger caps the total mass, replicators stop if they lose the Ledger's heartbeat, and an international kill code halts replication within one light delay.
- Copy errors that compound. A flaw copied every generation ends up in every machine. Blueprints are checksummed every generation, and each new machine is tested against its blueprint before it may build.
- A slow line. The loop runs at the pace of its slowest machine type. Lithocells are the likeliest bottleneck, which is why complex parts are held to of the mass and the Choir rebalances lines generation by generation.
- A slower doubling. The plan has no slack here. A longer doubling time delays the Lattice, cuts the energy it delivers before the deadline, and pushes the ice launches against the date after which ice cannot arrive in time (timeline).
- The environment. Mercury's day side is hot and its surface is bombarded. Polar crater rims avoid the worst of the heat, and the loop builds its own replacements.
Reproduce these numbers
The Foundry's numbers come from calc/src/model/foundry.ts, reading data/assumptions.yaml.
pnpm calc --only foundry-full-scale # landing + n doubling periods
pnpm calc --only foundry-power-per-kg # e ln 2 / T
pnpm calc --only veil-affordable # when one doubling's output equals the Veil
pnpm calc --only veil-diversion-days # growth the Veil costs at its close date
The assumption to try is the doubling time, foundry-doubling-days, in the
Explorer or with pnpm calc --set foundry-doubling-days=70. Lengthen it and
full scale, the Lattice and the first ice launch all slip; the energy budget goes negative, the Veil is no
longer affordable by its close date, and the simulated atmosphere misses the deadline. Shorten it and
everything moves earlier and the spare energy grows. The critical path
runs through this one number.