The six constraints
Technology does not advance because people become cleverer. It advances when a small number of physical limits are pushed back. Almost every step on the ladder is gated by one of these six, and knowing which one you are against tells you what to work on.
The usual way to tell the story of civilization is as a sequence of ideas: someone thinks of the wheel, someone thinks of the steam engine, someone thinks of the transistor. That story is close to useless if you actually have to rebuild any of it, because ideas were almost never the bottleneck.
What follows is the more useful framing. Six constraints, each of which caps what is possible regardless of how much anyone understands. When a step on the ladder feels impossible, it is nearly always because one of these is binding — and the fix is to attack the constraint, not to try the step harder.
What you can make is bounded by the hottest fire you can sustain and contain. Almost every materials breakthrough in history is a step on this one ladder, and the order is not negotiable: you cannot make cement before you can make lime, because lime is cooler.
The containment half matters as much as the heat. Reaching a temperature is easy; holding a vessel at it for hours without the vessel failing is the actual engineering problem, which is why firebrick and crucibles appear so early and matter so much.
| Temperature | What it unlocks | How you get there |
|---|---|---|
| ~600 °C | Cooking, water treatment, hardened wood, low-fired pottery | Open wood fire |
| ~900 °C | Lime, brick, common pottery, glass frit | Kiln with a chimney and draught |
| ~1,100 °C | Copper and bronze melting, iron bloom (solid state) | Charcoal + bellows |
| ~1,400 °C | Cement clinker, good glass, cast iron | Charcoal or coke + strong continuous blast |
| ~1,600 °C | Molten iron and steel in quantity | Coke blast furnace, preheated blast |
| ~2,000 °C+ | Refractories, specialty alloys, carbide | Electric arc |
| ~3,000 °C+ | Crystal growth, plasma processing | Abundant electricity |
The practical lesson: when a process fails, suspect fuel before you suspect ore. Groups that try to smelt with firewood conclude they have no metal, when what they lack is charcoal.
You cannot make a part more accurate than the machine that made it. This sounds like it should make progress impossible — and it very nearly does. The escape is a handful of self-referential tricks where accuracy is generated out of nothing but symmetry and patience.
The most important is the three-plate method. Lap two plates together and both can be curved — one convex, one concave — and they will still match perfectly. Bring in a third and lap all three pairings in rotation, and the only shape all three can share is a true plane. A flat surface, derived from no prior flat surface, using nothing but abrasive and time. Every machine tool in the world descends from that trick.
Screw threads escape the same way: a rough screw lapped against a long nut has its periodic errors averaged out, and each generation of screw is better than the one that cut it.
| Accuracy | Method | What it makes possible |
|---|---|---|
| ~1 mm | Hand, eye, simple tools | Carpentry, masonry, basic smithing |
| ~0.1 mm | Files, scrapers, surface plate | Working machine tools, clocks, instruments |
| ~0.01 mm | Lathe with leadscrew, micrometer | Steam engines, interchangeable parts |
| ~0.001 mm | Grinding, lapping, gauge blocks | Internal combustion, precision bearings |
| ~1 µm and below | Optical and interferometric methods | Optics, aerospace, instrumentation |
| nanometres | Photolithography | Integrated circuits |
The practical lesson: precision is infrastructure. It is invisible, it takes years, it produces nothing you can eat, and skipping it caps your civilization at tier 3 permanently. This is the most commonly skipped step in every rebuild plan ever written, including the one this site was built from.
Everything a society does that is not directly feeding itself is paid for out of surplus. The relevant number is how much energy you get back for the energy you spend getting it — often written as EROI, energy return on energy invested.
When surplus is low, everyone farms. As it rises, you can afford specialists, and specialists are what make technology accumulate. A blacksmith exists only because farmers produce more food than they eat.
| Source | Approximate return | What it supports |
|---|---|---|
| Foraging | low, highly variable | Small mobile bands, almost no specialists |
| Subsistence farming | a few to one | Villages, part-time craft |
| Farming with draft animals | higher | Towns, full-time trades |
| Water and wind power | high, but fixed in place | Mills, early industry |
| Coal | very high at first | Industrial revolution |
| Conventional oil | extremely high early, falling since | Mass mobility, modern agriculture |
| Modern renewables | moderate to high, and rising | Contested; depends heavily on storage |
A caution on these numbers: published EROI figures vary enormously depending on where the analyst draws the system boundary, and the field genuinely disagrees. Treat the ordering and the direction of travel as robust, and treat any specific ratio you see quoted — including in the sources on this site — as an argument rather than a measurement.
The practical lesson: the transition from current sunlight to stored sunlight at tier 4 is the single largest discontinuity on the ladder. Everything below it is capped by land area. Everything above it is capped by what you can dig up — and by what happens to the atmosphere when you burn it.
A skill needs a minimum number of practitioners to survive. Below that threshold it does not decline gracefully — it disappears, because the last competent person dies before finishing training a successor, and a partially-transmitted craft is not a craft.
The clearest documented case is Tasmania. Isolated by rising sea levels around 10,000 years ago, its population — of the order of a few thousand — appears to have progressively lost bone tools, cold-weather clothing and fishing over the following millennia. No catastrophe, no invasion, no forgetting in any single generation. Just a population too small to reliably keep every skill alive, losing them one at a time. (The interpretation is debated among archaeologists, but the pattern of technological simplification under isolation is well attested.)
The mechanism is worth being precise about: it is not that small groups are less intelligent. It is that transmission is lossy. Each generation learns slightly less than the best practitioner knew, and only a large enough pool of learners produces someone who exceeds the average and pulls the skill back up.
Rough thresholds — treat as orders of magnitude, not thresholds you can plan against:
- Tens — survival skills, basic craft. Every skill is one death from extinction.
- Hundreds — full-time specialists: a smith, a potter, a healer.
- Thousands — trades with apprentices and redundancy; a machine shop is sustainable.
- Tens of thousands — chemistry, medicine, engineering as distinct professions.
- Millions — the deep supply chains that semiconductors require.
The practical lesson: redundancy in people is not a luxury, it is the storage medium. Two people who know a skill is not twice as safe as one — it is the difference between a skill that exists and one that is a coin flip. This is why recording knowledge and teaching deliberately appear so early on this site.
You cannot run a process you cannot hold. This is the quiet constraint that decides the order of the chemical and thermal tiers, and it is almost always discovered the hard way.
- Heat needs refractories. Ordinary brick spalls apart at forge temperatures; your first furnace linings will fail until you make proper firebrick.
- Acid needs glass or lead. This is precisely why the lead chamber process for sulphuric acid was built out of lead, and why glassblowing gates all real chemistry.
- Pressure needs seamless vessels and good steel. This is why synthetic ammonia sits so high on the ladder — the chemistry was understood well before anyone could build a vessel that survived it.
- Vacuum needs glass, pumps and seals — which is why vacuum tubes cannot precede glassblowing.
- Purity needs containment that does not contaminate. At nine-nines purity, the container is a source of pollution.
The practical lesson: when planning any step, ask what holds it before asking how it works. The vessel is usually the harder half, and it is usually the half that is missing from the description in the book.
Knowledge decays. Not metaphorically — physically, in the medium, and socially, in the transmission. A civilization's technological level is not what it once knew; it is what it can still reliably reproduce this year.
| Medium | Realistic lifespan | Failure mode |
|---|---|---|
| Oral tradition | 1–3 generations | Drift, then loss on a single death |
| Fired clay | millennia | Breakage; very low density |
| Parchment | centuries to a millennium | Damp, fire, vermin |
| Rag paper | centuries | Fire, water |
| Wood-pulp paper | decades | Acid self-destruction; yellowing and crumbling |
| Magnetic media | years to decades | Decay, and readers going extinct |
| Optical discs | years to decades, highly variable | Layer degradation; readers going extinct |
| Flash storage | years unpowered | Charge leakage; silent corruption |
Note the shape of that table: the newest media are the least durable, and their real weakness is not the medium at all — it is that reading them requires a working civilization. A clay tablet needs eyes. A hard drive needs a power grid, an interface standard, a filesystem and a file format, any one of which can go extinct while the disk itself is still perfectly intact.
The practical lesson: store the critical subset redundantly, in multiple media, in multiple places, and include at least one form that needs no technology to read. Print the fifty manuals that matter. See Take it offline for the actual plan.