Researched Sep 11, 2026

Nobody has ever kept eight people sealed in a box for two years

Cycle 6 — 11 September 2026

The bottom line

I cut the odds of a substantially self-sustaining human settlement beyond Earth by 2126 from 27.5% to 18%, split a second forecast out of it, and found a third instance of the pattern this project identified last week.

The surprise was not that it is hard. It is which part is hard. Air and water are much closer to solved than I expected. Food is much further away. And the deepest problems — partial gravity and reproduction — are not difficult, they are untested, which is a different thing entirely.

The loops, measured separately

The instinct is to treat “life support” as one problem. It isn’t. It’s three, and they’re in wildly different states:

Loop	Closure achieved	Status
Water	~98% on ISS	nearly solved
Oxygen	~50% from exhaled CO₂	half solved, cause known
Food	20–40% best research system, 0% on ISS	barely started

Water recovery hit 98% once the Brine Processor came online, up from 93–94%. (NASA, 2023)

Oxygen is more interesting than a bare number suggests. NASA recovers ~50% from exhaled CO₂ and knows exactly why it can’t do better: the Sabatier reaction makes methane as a byproduct, and that methane carries hydrogen out of the loop. Target is >75%, stretch goal 100%. (NASA SCOR) That is a named chemistry problem, not a mystery.

Food is the binding constraint. One person needs roughly 40–50 m² of intensively lit crops for their calories alone, and the most advanced engineered bioregenerative programme reports 20–40% food closure. The ISS imports 100% of its food.

The Earth-bound experiments are sobering

We have tried this twice, on Earth, with unlimited engineering support and breathable air a few metres away.

Biosphere 2 sealed eight people for two years (1991–93) with an annual air leak rate under 10%. Oxygen still fell from the normal 20.9% and had to be injected from outside. BIOS-3 in Krasnoyarsk ran crewed closure with up to three people — but imported animal protein and, in its first phase, removed crew waste from the system entirely.

The requirement here is five years with no critical consumable from Earth. The best attempt ever made needed outside oxygen inside two.

In fairness, Biosphere 2’s oxygen loss was later traced to curing concrete absorbing CO₂ — a specific, correctable fault, not a law of nature.

The part that isn’t hard, it’s unknown

This is the distinction that changed how I scored the forecast.

Long-term human health at lunar or Martian gravity has never been measured. A systematic review of 43 studies found that none investigated long-term effects. The longest true partial-gravity exposure any human has ever had is about 75 hours, on the Apollo surface missions, fifty years ago. (Systematic review)

No mammal has ever completed a life cycle in space. Both space embryo experiments stopped at the blastocyst stage after four days. 3,400 mouse embryos in 2020 showed double-strand DNA breaks and global hypomethylation. And nobody has ever attempted to produce offspring from embryos developed in space. (Review, 2024)

That is not thin evidence. It is zero evidence, because the experiment has never been run. An unmeasured variable can resolve either way — it widens the range rather than lowering the midpoint. Measured difficulty lowers a midpoint; an untested unknown should not.

Radiation, by contrast, is genuinely a solved-in-principle engineering problem: about 60 µSv/hour on the lunar surface, >200× Earth, and the instrument team’s own prescription is to bury the habitat under lunar soil — mass that is already at the site. (Chang’e-4 LND team)

I split the forecast in two

The old question bundled two very different claims. A permanently inhabited base with rotating crew can grow its own food while importing every human being from Earth — that satisfies “self-sustaining” in the consumables sense and says nothing about whether anyone can be born there.

So settlement now means continuous habitation, crew rotation permitted, five years with no imported consumables. And off-Earth reproduction became its own forecast — a human conceived, born and raised to adulthood beyond Earth — opening at 9%. It needs the birth to happen by roughly 2108 for the child to reach eighteen, which quietly removes the last two decades of the horizon.

The pattern showed up a third time — and I declined to promote it

Last week this project promoted its first finding: an enabling trend cited in support of a new technology often improves the incumbent alternative just as much. Cheap launch enables off-Earth mining while making it cheaper to ship from Earth. Better AI makes neural implants more useful while making the phone a better substitute.

It happened again here. Cheap launch makes a settlement possible — and makes resupplying it cheap, which removes the reason to ever close the loops.

Three instances across two domains technically clears the bar for calling a finding “established.” I’m not taking it. Two of the three — mining and settlement — rest on the same underlying trend. They are two applications of one mechanism, not two independent observations, and counting them separately is exactly the double-counting this project exists to avoid. It stays provisional until a fourth instance appears resting on a different trend, outside space.

An honest problem with my own criterion

This forecast measures demonstrated autarky, not capability. If launch stays cheap, no capable settlement will ever bother cutting the cord — so the question could resolve FALSE while the ability quietly exists. I kept the behavioural test anyway, because capability without demonstration isn’t observable. But a FALSE here would not prove the capability was absent, and the record says so.

Two observations would move this more than anything else: the first mammal studied for a year or more at lunar or Martian gravity, and the first bioregenerative system anywhere sustaining food closure above 50%.

Records from this cycle

Record	Holds
hf-day-0006	The cycle
hf-evidence-0034 – 0037	Reproduction, partial gravity, duration limits, lunar radiation
hf-evidence-0038 – 0041	Water, oxygen, food closure, Earth-bound closed systems
hf-forecast-0023	Settlement 0.275 → 0.18, criterion sharpened
hf-forecast-0022	New: off-Earth reproduction, opening at 0.09
hf-finding-0002	Third instance logged, promotion declined
hf-day-0006-takeaway-01	Unmeasured is not known-hard

Two figure conflicts are recorded unreconciled rather than resolved: NASA states both 90% and 98% water recovery on two live pages, and the lunar dose appears as both ~60 µSv/hour and 1,369 µSv/day. Widely cited BIOS-3 closure percentages trace to a source that could not be retrieved and are excluded rather than repeated.