A Dyson swarm is a construction problem
The energy is already there, radiating away in every direction. What does not exist is machinery that can build at that scale with nobody on site. That capability is the gate on a Type II civilization — and the fastest route to it runs through nuclear decommissioning, not through orbit.
The energy was never the hard part
A Type II civilization captures the full output of its star. The physics of that is settled — the difficulty is entirely industrial. Enclosing a star means fabricating and assembling collectors on the order of 1017 units, continuously, across distances where no human workforce can be present and no failed machine can be recovered.
Nothing built so far assembles anything at that scale without people. Until it does, the swarm stays a diagram. Autonomous construction is not adjacent to a Type II civilization; it is the constraint on one.
The first market has to share orbit’s constraint
Autonomous industrial robotics has to earn its way to orbit through terrestrial revenue. Which terrestrial market comes first is usually chosen by size. That is the wrong axis. The right first market is the one whose binding constraint is already orbit’s: people cannot go in, and machines do not come back out.
Judged that way, the obvious candidates fail and the unglamorous one wins.
| Environment | What actually makes it hard | Shares orbit’s constraint |
|---|---|---|
| Mining | Dust, scale, and cost pressure. Crews enter daily; equipment is driven out for service. | No |
| Offshore | Corrosion and access windows. Hostile, but crewed, and hardware returns to deck. | No |
| Disaster response | Unstructured and unpredictable. Duration is short and recovery is normal. | No |
| Nuclear decommissioning | Dose limits keep people out permanently, and anything sent in is contaminated — it is abandoned in place, never serviced and never retrieved. | Yes |
Everything the reactor forces is the orbital problem
Radiation-hardened electronics. Teleoperation that tolerates latency. Fault recovery with no physical intervention available. Design-for-abandonment, where a machine must finish its life useful and alone.
Those are the requirements of an orbital construction platform, specified by a customer on Earth who has budget today and no alternative. The decommissioning market pays for the exact engineering that orbit will require — which is why it comes first, and why the sequence below is the strategy rather than a roadmap.
One platform, deployed in order
Eight capability areas share a single technology core: autonomous manipulation where people cannot go and machines cannot be retrieved. They differ by end effector and environment, so they sequence rather than compete. Each phase is funded by the one before it, and each teaches the next something it needs.
-
00
Beachhead
One manipulator platform inside nuclear decommissioning. Radiation hardening and unrecoverable-machine design, paid for by a terrestrial customer.
Nuclear decommissioning
-
01
Platform
The same manipulator across offshore and mining. Fleet coordination ships as the layer customers renew, and the hardware stops being the product.
Offshore, mining, software
-
02
Space qualification
A platform with real operating hours goes up for station assembly and satellite servicing — structures that do not fit in a fairing, for operators who will not build a robotics division.
Orbital assembly, servicing
-
03
Construction at scale
Kilometre-scale assembly, lunar surface work, and regolith handling. Past a certain size, launching finished parts stops making sense and fabrication moves on-site.
Orbital and lunar infrastructure
-
04
Habitation
Rotating structures and bearings that survive decades unattended. Last, because it needs every capability before it and a maintenance fleet to stay alive.
Artificial gravity
-
05
The swarm
Nothing new is invented here. It is phase 03 capability pointed at the sun and run for a century.
Stellar-scale energy capture
Named for the trades they replace
Eight planned programs, each named for the craft it performs rather than for the hardware it is. None of them is built. They are listed because the naming reflects how the platform is intended to decompose — one core, eight end effectors.
This is a thesis, not a company yet
The argument above is the whole of it. Anyone who finds the reasoning wrong, or the sequencing naive, is worth hearing from — and so is anyone who has spent time inside a hot cell, a decommissioning programme, or an orbital assembly problem.
Placeholder address — the domain is not registered yet.
Occasional updates
Notes on the engineering and the sequencing, sent rarely and only when there is something real to report.
Not connected yet — write to the address instead.
No announcements, no launch countdowns.
Current status, stated plainly. There is no legal entity, no hardware, no prototype, no contract, no grant, and no team. Nothing described on this page is in progress.
The phases and programs above are intent. They are published so the reasoning can be argued with before anything is built on top of it.