15. Less-Dumb Requirements
Exploration was the initial dream for Mars, independence the ultimate one. You can taste the difference in the air.
The first three years on Mars answered exploration cleanly: pressurized rovers, pop-up hab tents, small inflatables, a hub to dock and debrief. Vehicle-based field science works. It is honest, brave, and — by design — temporary.
But exploration is a postcard, not a postcode.
In the first century of settling Mars for a proper sized civilization, what mattered was not the age of footfalls; it was the age of foundations.
To describe that work and set of requirements, we have to step back from the romance of firsts and ask duller questions with larger answers.
♦
Start with ends, not beginnings. The target was a self-sustaining city on Mars.
More technically a Independence Level 7 community (note: please recap the Prelude Mars Codex — Insurance & Independence — if you don’t know what that means). In other words, a population large enough to carry its own knowledge loops — education, medicine, repair, agriculture, manufacturing — without begging Earth for every spare gasket or surgeon.
To survive, if the ships from Earth stopped coming. In practice we learned to treat 10,000 people as the floor, not the ceiling.
What does 10,000 people — and much more — mean in steel and air?
Earth offers analogs. A hotel room is ~22 m² per guest; a cruise-ship cabin ~18 m² (double occupancy assumed); a modest New York apartment ~64 m² for one. Martian private quarters lived somewhere in that band — call it 22–36 m² per city resident in the founding century, even then we were efficient but still human. We knew nobody was going to live on Mars if it was like a life-sentence in an army barracks or worse, a submarine bunk.
What’s more, private accommodation is not the city. A city’s space is private + shared + working + civic + the corridors between. On Mars the multiplier was unglamorous and non-negotiable: roughly ×3 to x5 to carry life support, clinics, teaching rooms, kitchens, green halls, workshops, storage, and link ‘streets’. And higher multipliers later when we needed to add the secondary industries to really achieve the IL-7+ goal of ‘self-sustaining.’
Do the arithmetic:
10,000 people × 22–36 m² private = 220,000–360,000 m²
Now ×3-5 for everything shared = >660,000–1,080,000+ m² pressurized, shirtsleeve territory.
That is one to two Pentagons of floor area (≈ 620,000 m²), three to four Empire State Buildings (≈ 254,000 m² each), or ~2–5× the deck area of an early 21st century ocean megaship (Icon-class ≈ 240,000 m²).
To build a (small self-sustaining) “city” on Mars, we were, in effect, building a bungalow cruise ship in a desert that hates ships — and then as the city grows promising to extend that desert ship a block a month for decades.
Of course our MarsX plan numbers were far more precise than this as you will see, but what matters here is the scale.
Scale first. Elegance later.
♦
From this scale flowed an obvious, often ignored requirement: the city must be lived mostly inside, not between suits. Every settlement designer in the early twenty-first century on Earth seemed to miss this, or at least not fully embrace the implications at a scale of 10,000 people. Obvious, though it is now to us Martians with hindself.
In films and sci-fi, EHA — Extra-Habitat Activity — is freedom. On Mars and in reality, EHA is friction. A conservative don/doff cycle is ~30min each way; suit joints grind dust; seals age; filters saturate; an airlock queue burns useful hours as surely as a generator burns methane. And let’s not even talk about all the heavy complex dirty-clean door interfaces you would need everywhere — even with the much better door-design approach than those twenty-first century monster plans, doors were the bain of my building life in that first century. The bottom-line, if your residents must suit up to visit a neighbor, you have built a museum, not a municipality.
The only survivable default is to build a Mars life in shirtsleeves: large, continuous pressure graphs with isolation bulkheads and sane egress. Fault containment lives at every junction; pressure doors make neighborhoods; maintenance aisles are streets; transit happens at one atmosphere (or thereabouts). You pay the mass to enclose space once and then amortize it with use.
And yes, the outside still exists — mines, pads, farms under cover. All Martian’s of course regularly don their EHA suits, but then the reason is to go outside and do something outside, not go outside just to get inside again.
So on Mars once the building was done, the design was for a steady-state where most life, work, and most logistics belonged inside, behind two skins and a thousand valves.
Dust stays outside. People, mostly, do not.
♦
Shipping from Earth was generous given the twenty-first reusable rocket solutions, but it was never going to be enough. Even with a heroic thousand-ship per hoff (Hohmann liftoff window) launch cadence, we learned that mass, manufacturing, and maintenance beat manifests.
Mass: every kilogram delivered must buy the maximum m² of pressurized utility. The structures that won were those whose shipped parts were thin, simple, and repetitive — interfaces, liners, frames, manifolds—and whose bulk came from Mars itself.
Manufacturing: design for a factory we could land and then grow. That meant SKU discipline ruthless enough to be boring. Fewer geometries, fewer gaskets, fewer fasteners. The parts catalog reads like a short poem; the yard repeats it until the city appears.
Maintenance: nothing glued into walls we cannot reach; everything swappable in gloves; software states legible in a bad hour. If a tired human with a wrench and a bot with one good arm cannot service it, it is not a Martian system. Design to operate and fix.
ISRU (In Situ Resource Utilization) was not a flourish; it was the load path. Regolith became shell and shield; water became wall and loop; shipped materials reserved for precision and pressure. The ratio drifted over time, but the direction did not.
♦
Iteration frightened Earth more than Mars. Earth loved monuments; Mars loved versions.
We did not get it right the first time. Or the second. The trick was to design so that later versions could arrive without funerals of earlier ones. Keep the form-factors, philosophy and linkage steady, evolve the internals brutally. I watched Earth learn this lesson with rockets: change engines, tiles, welds, software — keep the 9m core, methane-fuel, and reuse philosophy to keep the mission. The ship that flew in Year 5 and the ship that flew in Year 25 looked like cousins and performed like different species.
Our buildings behaved the same. We kept the two or three recurring shapes that production liked, then changed everything inside them as power grew, printers matured, and our bad ideas failed loudly. The city stayed recognizable enough to extend and strange enough to improve.
“Repetition over perfection” read like defeat in slide decks. In yards, it read like progress.
♦
From these first settlement design principles we minted a handful of soundbite objectives — rude enough to argue with, sharp enough to steer by:
Design for thousands, not tens. If scaling by ×10 breaks the layout, the layout is wrong.
Shirtsleeves by default. EVA is an exception, not a commute.
Platform, not portfolio. One pressure port, one corridor section, one utility manifold family.
Factory before flag. Power and robots sized for throughput, not survival (and posters.)
ISRU mindset from Day 0. Let Mars carry Mars; ship only what Mars cannot yet make.
Extreme iteration assumed. No tombstones; everything refittable; versions expected.
Tool simplicity. The wrench you can land once and use for decades is worth more than a clever clip you must reorder.
Measure what compounds. m² per tonne landed; % robotic assembly; human-hours per 100 m²; standardization index; mean time-to-swap.
These were not slogans once we printed them on clipboards. They were the test that drawings had to pass before the yard touched dirt.
♦
There were, equally, anti-requirements — habits we had to break:
No bespoke showpieces that cannot be copied blind.
No layouts that require heroics to add the tenth unit.
No systems hidden so deeply you must cut them out to touch them.
No reliance on perfect geology or perfect weather.
No comfort-first interiors that sabotage maintenance.
No assumption that Earth will keep shipping what we forget to learn.
If a proposal depended on any of the above, we filed it under “pretty” and built something else.
And so these were the principles and foundations that the “MarsX (Direct) Plan” was built upon. The detailed operational and evolving plan that took Zubrin’s vision and started to make it concrete. Literally.
♦
By the time the numbers sobered us — hundreds of thousands of square meters under pressure; years of continuous build; megawatts not for people but for factories, road-trains and robots — we had learned the only lesson that ever mattered: you do not build a settlement; you build the system that builds the settlement, and you live in the spaces it leaves behind.
That shift did not fit on a poster. It did, eventually, fit in a city.
And yes, I was there — on my droid inspection rounds with a tablet and two drones, in airlocks that hissed like kettles, at meetings where someone always asked, “Could we make this smaller?” and someone else replied, “Only if you want to rebuild it next year.” We were not heroic. We were, at our best, competent and consistent. That turned out to be enough.
♦
We made the principles operational by turning them into metrics. Targets are not poetry; they are culture. I remember one from my early years in the third decade HE, which every morning board carried the same five:
Areal Efficiency (AE): m² pressurized per tonne landed.
Robotic Assembly Ratio (RAR): % of steps completed without humans.
Human-Hour Install (HHI): crew-hours per 100 m² to handover.
Standardization Index (SI): % of parts drawn from the common catalog.
Mean Time To Isolation (MTTI): minutes (ideally seconds) to seal a faulted segment.
Targets were blunt: AE > 50 m²/t by Gen-3; RAR > 90%; HHI < 4 h/100 m²; SI > 70%; MTTI < 5 min. If a drawing could not plausibly improve at least two of these without worsening the others, it did not reach the yard. If a field change did not move a counter in a week, it reverted. I have seen Earth careers rescued by a rising SI and ended by a stubborn HHI.
Metrics are kind. They do not care who proposed the idea. They only care whether we will still be alive next winter.
♦
Takt-time arrived the day we stopped “building a base” and began running cells.
A cell was a rectangle of ground and a slice of schedule. Its script was always the same:
grade → lay trunk → place shell → berm/print → install liner → route utilities → pressure test → handover.
Grade meant sub-5 cm over 30 m. Trunk meant power/fluids/data manifold with pre-terminated nodes at 10 m spacing. Place shell meant lift, align, clamp, leak-precheck. Berm/print meant ≥0.5–2 m equivalent shielding mass by robot only. Liner meant snap-fit, gland, and first air at 0.3 bar. Route utilities meant plug the ladder to the trunk and verify telemetry. Pressure test meant creep to 0.7 bar, soak 4 h, sniff for mic, escalate to 1.0 bar. Handover meant the board ticked over and the next cell started before anyone finished a sentence.
Cells ran in parallel. When one printer choked, the yard flexed — haulers diverted, cranes re-tasked, inspectors reassigned. Throughput, not heroics. A good week was five handovers per cell. A great week was when no one noticed.
♦
Utilities were the spine. Streets before houses, always.
Power first: ring mains with sectionalizers every 30 m, dual-fed, sized for factories not kitchens. Fluids next: water, process, waste, each with color-coded quick-couplers and visible pressure gauges you could read through dust. Air last: not because it mattered least, but because air lived inside and streets lived outside. Data rode all of it, redundant, boring, fast. We collectively called this ‘Power & HWAC’ — as on Mars water meant more than ventilation.
The sequence never changed: power → robots → shells → liners → life support. Without the first two, the rest are theatre. We learned to stage megawatt-scale banks before the first hab was habitable, because shells and robots eat energy even when people do not. The temptation to rush a cozy interior ahead of a sober trunk was strong; we buried that temptation under conduit and moved on.
Bulkheads every segment. Isolation valves at every knee. Utility ladders at the same pitch everywhere they appeared. When fire codes and good sense argued, we sided with both: wider aisles and more bulkheads. The graph stayed connected even when a careless bot tore a line at 03:17. We cursed, sealed, rerouted, and the counters kept time.
♦
We formalized gates so arguments would end quickly:
Gate A: Ground & Power. Trunk energized to spec; backup path verified; thermal rejection online.
Gate B: Shell & Shield. Shell placed, joined, and shield mass verified by lidar + scale; no human-placed buckets counted.
Gate C: Liner & Leak. Liner installed; pressure creep passed; acoustic/UT scan clean.
Gate D: Utilities & Life. Manifolds plugged; ECLSS cartridges mounted; alarms latched; failover tested.
Gate E: Handover. The two numbers that mattered — HHI and MTTI — posted to the board, with the delta from last week.
No “soft opens.” No “almosts.” The city is not a restaurant.
On a good night, Gate C sounded like kettles: small hisses, then silence. On a bad night, Gate C sounded like opinions. Bulkheads closed; repair kits arrived; the clock did not stop.
♦
Iteration worked because we froze the right things and abused the rest.
We froze: pressure port geometry; corridor cross-section; utility ladder interface; bulkhead actuation spec; airlock vestibule clearances; aisle widths. Change any of those and your SI collapses.
We abused: printer heads and mix ratios; crane end-effectors; bot fleet composition; liner panel layups; software states and HMI; berm compaction technique; heat-recovery routing. On Tuesols we ran A/B tests like a habit: two cells, same drawing, different nozzle diameters or autonomous pathing; winner keeps its job. “Winner” meant AE up, HHI down, or RAR up with no safety regressions.
Some changes were rude and right. We widened all aisles by 0.4m after one quarter when average HHI refused to drop; spares ran faster, arguments stopped. We standardized a single gland style after a month of incompatible O-rings; SI leapt 18 points; leaks fell by half. We killed a clever integrated rack because you needed three hands to service it. The designer sulked; the city purred.
♦
The anti-requirements saved us most when we were tired.
We nearly stranded a plaza under glass because a committee loved the render. The plaza murdered turning radii, pinched the haul lanes, and blocked a crane arc. Counters told the truth: AE dipped; HHI spiked; incidents rose. We cut the plaza into ribs and let light in between. The render looked worse. The yard ran better.
We nearly lost a winter to a “smart wall” that hid life support inside composite panels. Beautiful. Non-serviceable. On the third frozen quick-disconnect we tore the panels out and replaced them with rail-mounted cartridges and honest pipes. HHI halved. MTTI fell under three minutes. Pride took longer. Frontline engineers smiled.
We once waited eight months for a lava-tube survey that never returned a site we could reach with the fleet we had. Those months built nothing. We stopped waiting for geology and built where the printers could walk.
None of these were dramatic. They were the accretion of adult decisions.
♦
Human construction crew time was the currency we never printed enough of. We protected it with pre-terminated everything.
Pre-wired manifolds arrived in sleeves; labels were embossed and color-fast; QR tags told tired eyes what the human brain would forget at 02:00. We learned that moving from “cut to fit” to “fit to click” is the difference between 9 h/100 m² and 3.5 h/100 m². The latter keeps music in the mess on Thursol nights.
Swap discipline won arguments no paper could. Every cartridge — filters, pumps, CO₂ beds, scrub fans — had a clock target: two-person, <30 min, in gloves, standing room. If you needed a third person, you redesigned the bay. If a software state could not be read on a dumb terminal at the point of use, it did not exist.
There were weeks we hit targets and morale anyway felt thin. We added viewports that were not windows: periscope shafts with extra mass where it mattered. HHI did not change. People did. Not every counter fits on a board.
♦
By the mid founding century (human era) the metrics stabilized at grown-up values; the yard’s choreography went quiet. You can tell a mature system when meetings grow shorter and forklifts move like memory.
From the air (and yes, sometimes I took the drone up for pleasure) the pattern was obvious: trunks first, rings of protected passage, large volumes attached where utilities smiled and robots could stage, knees at predictable intervals, bulkheads like punctuation marks in a long, careful sentence. We had not designed a masterpiece. We had learned a grammar.
What exactly did that grammar write in the first century after the first decades sprint — the thing whose outlines you can now recite from memory if you’ve lived under it?
I will not spoil it here. Only this: when frontline construction site-reliability engineers were asked to name the two repeatable form-factors — the civil-engineering equivalent of “ship” and “booster” that made the settlement mass math sing — they did not choose poetry.
They called them “barrels and barns.”
