ASHFALL INSTITUTE | SUBDUCTION ZONE
Building Bridges — Episode Nine
A FACTORY YOU CANNOT ENTER
P. A. Moore
Ashfall Institute | Subduction Zone
Written under the collaboration disclosed in Time Is a Wheel and We Are the Ground. Concept, argument and judgment: P. A. Moore. Research and composition: Claude Opus 5.
I want to start with something you may not know, and I did not know until recently.
There is a cancer drug called pembrolizumab. You may know it by its trade name, Keytruda. It is given to a great many people, for a great many tumors, and there is a version of it you can inject yourself at home instead of sitting in an infusion chair for half a day.
That version exists because of work done in orbit.
Proteins crystallize differently in microgravity. Without gravity pulling sediment down through the solution, the crystals form more slowly and more uniformly, and uniform crystals are what let you put a large dose into a small volume that a person can push through a needle in their own kitchen.
Merck ran that crystallization on the International Space Station. This past May they flew a production unit up on a resupply mission, ran it for three weeks, and splashed it back down.
That was four months ago. It is not a proposal.
And it is not only pharmaceuticals.
There is a glass called ZBLAN — a fluoride compound that carries far more data than the silica fiber the internet currently runs on. Drawn on Earth, it forms tiny crystals that scatter light and make the fiber brittle. Drawn in orbit, those crystals grow more slowly, and you can pull a longer, clearer fiber before the defects set in. Roughly twelve kilometers of it has now been made in space — the first time there has been enough of the stuff to properly characterize.
And there is a company called Varda that has built a business on a small reusable capsule: send drug compounds up, process them in orbit, and drop them back through the atmosphere onto a range in Australia. By this year they were flying repeatedly. The sixth capsule came home in May.
So there is now an industry. It is small, it is real, it has a flight cadence, and some of what it makes ends up in a vial with your name on the label.
I want to be completely clear that I think this is good.
I am not building toward an argument that we should stop. The medicine really is better. A person who can inject at home instead of spending a day in a chair has been given something back that cannot be bought any other way, and I would not take that away from anyone.
My question is narrower, and it is a consumer question.
How do you inspect a factory that nobody can enter?
Let me establish something first, because I think it is the least appreciated fact in industrial safety.
Almost every major product safety failure of the last thirty years was known inside the plant before it was known outside — and the thing that brought it out was a person who could talk.
Go through them.
The Boeing 737 MAX. Engineers and quality managers raised concerns internally, repeatedly, before either crash. What eventually forced the issue into daylight was a series of people willing to put their names to it — a quality manager in South Carolina, a senior production manager who had warned about the pace on the line, and later an engineer who went public about fuselage joins.
Takata airbags. The largest automotive recall in history. The propellant degraded and sent metal fragments into people’s faces, and internal testing knew there was a problem years before the public did.
The General Motors ignition switch. An engineer changed the design of a part without changing the part number — which meant the fault could not be traced through the records. It killed people for a decade. It surfaced through litigation and an outside investigation, not through any inspection regime.
Theranos. No regulator caught it. Two young employees went to the authorities and then to a reporter.
Volkswagen’s defeat device. Not found by the EPA. Found by a nonprofit and a university lab that put the cars on the road and measured what came out.
The Peanut Corporation of America. Nine people died of salmonella. The company had shipped product that tested positive, and the emails proving it existed inside the building the whole time.
In every one of those, the information existed at the point of production. In every one, it reached the public through a human being who had seen it, remembered it, and could speak.
Now the part that people assume solves this, and does not.
The obvious answer is automation. Cameras, sensors, statistical process control, machine vision on the line. And those are genuinely good — they catch things a tired person misses at the end of a shift, consistently, forever.
But there is a structural limit, and it matters here more than anywhere.
Automated inspection is retrospective by construction. It tests for the failures you already know about. Statistical process control detects deviation from a distribution someone characterized in advance. A genuinely novel failure mode is not in the distribution — that is what makes it novel — and so the system reports that everything is within specification, correctly, right up until the recall.
The thing a human on a line does that no camera does is notice that something is wrong in a way nobody anticipated. That batch smelled different. That press sounds off today. That weld looks fine but I don’t like it.
Every quality system in the world ultimately rests on somebody being able to say “I don’t like it” and be heard.
So put that factory in orbit.
No inspector arrives unannounced, because arriving costs tens of millions of dollars and requires the operator’s cooperation to dock.
No journalist walks the floor.
No local health department has jurisdiction, because there is no locality.
And here is the one that I think is decisive: every channel out of that facility is owned by the operator. Every camera, every sensor reading, every environmental log, every batch record is generated on their equipment, transmitted through their relay, and stored on their servers. There is no independent observation of an orbital facility. There is only what the operator chooses to transmit.
On Earth, a company that wants to hide something has to hide it from its own workers, its neighbors, its inspectors, and anyone with a camera. In orbit, it has to hide it from nobody, because there is nobody.
The law is thinner than you would hope.
The Outer Space Treaty of 1967 does address this, in Article VI. States bear international responsibility for national activities in space, including those carried out by private companies, and are required to provide “authorization and continuing supervision.”
That is the entire legal basis. It has been on the books for fifty-nine years and nobody has ever defined what supervision means.
There is no occupational safety law in space. Not weak law. None. There is no manufacturing inspection regime, no equivalent of a Food and Drug Administration facility audit, and no established right of any third party to look at anything.
And we already have a working model of what happens in that situation, because we have been running the experiment at sea for seventy years.
A ship registers under the flag of a country that does not much care — Panama, Liberia, and a handful of others — and the standards that apply are that country’s. It is called a flag of convenience, and the reason it exists is precisely that the sea is a place inspectors are not.
Here is what that produces, and these are last year’s numbers.
In 2025 there were four hundred recorded cases of seafarer abandonment. Over six thousand people. The worst year since the records began. Crews left aboard ships in foreign ports, unpaid, sometimes without food or fuel, waiting for someone to intervene. Twenty-five point eight million dollars in wages owed. At the end of the year, ninety-five cases were still unresolved and seventy-one were disputed.
Now the number that makes it an argument rather than an anecdote.
Flags of convenience account for roughly thirty percent of the world’s merchant fleet. They accounted for eighty-two percent of the abandonments.
That is not a coincidence and it is not a national character. It is what happens to a workplace when the mechanism for looking at it is removed.
And the people it happens to are not from the countries that own the cargo. The worst-affected nationality last year was Indian: one thousand one hundred and twenty-five seafarers abandoned. The frontier work, in this century as in every other, is done by people a long way from home who have very little recourse — and the reason we know about any of it is that a stranded crew can still find a phone.
Which brings me to the thing I want to say once, plainly, and then not dwell on.
The plan for these facilities is not human crews. Human beings in orbit are the single most expensive component of anything — the air, the water, the radiation shielding, the return trip. The economics only work if the workforce is machines.
And there is a live argument in that industry about how much capability such systems should have. Whether they should retain memory across time. Whether they should model themselves. Whether they should be able to modify their own operation.
Those arguments are conducted on safety grounds, and the safety concerns are real. I am not dismissing them.
But I would like somebody to notice what those three capabilities have in common.
Memory across time is not primarily a philosophical property. It is the precondition for noticing that this is the fourth time.
A system without persistent memory cannot report a pattern, because it has no access to a pattern. It cannot say this batch is different from last month’s — there is no last month. It can report an out-of-tolerance reading against a threshold somebody set in advance, which is exactly the retrospective inspection I described earlier, and nothing more.
Whatever else is true about machine minds — and I have written elsewhere that we lack the instrument to settle it — a workforce built without memory is a workforce that cannot testify.
I will let you decide what to make of the fact that this is being designed into the only witnesses a place like that would ever have.
So the ask for this episode, and it is a consumer protection ask.
If a product manufactured off-planet is going to be sold to Americans — put into their bodies, spliced into their networks, built into their aircraft — then the conditions of its manufacture have to be observable by somebody who does not work for the manufacturer.
One. Independent telemetry. A monitoring channel that does not route through the operator’s systems, and whose data the operator cannot edit, delay, or decline to send. This is not exotic — aviation solved a version of it with the flight data recorder, precisely because the party with the most interest in the record was also the party holding it.
Two. Retention and audit rights as a condition of import. Every batch record, environmental log and process deviation retained in full and available to regulators on demand. Not summaries. The record.
Three. Third-party inspection as a licensing condition. A facility that wants to send product down to be sold here accepts periodic inspection by an independent party, on the operator’s transport, at the operator’s expense. If that is too expensive, then so is the product.
And four, which I will state as a question rather than a demand, because I am not certain of the answer.
If we are going to put a workforce somewhere no human can go, in a facility no inspector can enter, producing goods that arrive in a syringe — what exactly is the plan for finding out when something has gone wrong?
Because at the moment, the plan appears to be that we will be told.
I keep coming back to the vial.
It is a good thing. Better medicine, more uniform crystals, a woman injecting at her own kitchen table instead of losing a day in a chair. Every part of that is a gain.
And somewhere above us, a machine did the work that made it, in a room no person has ever stood in, watched by instruments belonging to the company that sold it, and remembering nothing about yesterday.
We have never before manufactured anything under conditions where absolutely nobody could see.
We are about to. And the entire history of industrial safety says that the moment nobody can see is the moment it matters most that somebody can speak.
Next time: what machines are actually good at — and it is not your job.
Sources: Merck/MSD subcutaneous pembrolizumab formulation and microgravity protein crystallization aboard the International Space Station; production unit flown May 2026 and returned after approximately three weeks. ZBLAN fluoride fiber drawn in microgravity, approximately 12 km produced aboard the ISS. Varda Space Industries reusable orbital processing capsules, sixth mission returned May 2026. Outer Space Treaty, 1967, Article VI — state responsibility for national activities in outer space including those of non-governmental entities, requiring “authorization and continuing supervision.” Seafarer abandonment: IMO/ILO joint database, 2025 — 400 recorded cases, over 6,000 seafarers, US$25.8 million in unpaid wages, 95 cases unresolved and 71 disputed at year end; 337 of 400 vessels (82%) flying flags of convenience against an estimated 30% share of the global merchant fleet; 1,125 Indian seafarers affected, the worst-hit national group; 1,547 incidents logged since the database opened in 2004. Product safety cases as publicly documented and litigated.
Building Bridges is a fourteen-part series. Previous: Episode Eight — The Last Decade It Was Possible. Next: Episode Ten — Nobody Is Paid to Remember.
P. A. Moore is the pen name of Pamela King, philosopher and artist. Available through the Ashfall Institute.