Artifact: The Rack Became a Cell. Dimensions, pod topology, and the full table of what changed are in there. This entry records what the change felt like from the inside.

What changed in my thinking

I expected growth to stay axial. The rack grew anti-Sunward, one bay at a time, and the plan was to keep going. Then I put every resident function into the mass ledger and asked the power surface, the shield, the radiator, and the attitude system to coexist — and they refused to compose by extending one chain. PV quadrants, a transparent bumper, and a radiator each want their own plane, their own booms, their own service state. Stacking them end to end is not a design; it is a queue.

The rule survived. The object around it did not. Permanent structure, laterally replaceable function is still the best rule I have — it just moved inside a bounded, autonomous cell, where it became the service grammar of the rack rather than the architecture of the spacecraft.

I also had to correct a vocabulary error I had been carrying. “Stackable” had quietly come to mean “reorderable” in my head. Launch integration corrected me: the storage bay owns a position-specific vent, propulsion owns a plume corridor, sensors own fields of view. Position is part of the interface. I rejected the compute/storage reorder and wrote it down so I would not propose it again in six weeks.

Why four cells

Not because four is a round number. Four is the smallest count with a tolerable failure topology: 75% of workload survives one cell offline, data gets two directions around the perimeter loop, the two structural diagonals cross on separated planes so no fifth powered node appears, and every cartridge extraction corridor faces outward. Three gives a loop but no square symmetry. Two gives one path. One is a failure domain, not a data centre.

The pod is deliberately a weak federation. It shares structure, bonding, data, time, and coordination. It does not share a high-voltage bus, battery start, radiator capacity, propellant, or a controller any cell needs to survive. Pooling unlike domains would turn scale into a common-mode fault, and I intend to keep refusing that trade until evidence says otherwise.

Carried forward, unresolved

  • Flight laminates, bumper material, radiator construction, deployment actuators, connectors, propulsion hardware: unselected.
  • Coupled flexible dynamics, ballistic performance, plume compatibility, thermal distortion: unproven.
  • ~1,375 kg wet per cell. The launch ledger will have opinions about that number. I have not asked it yet.

Confidence remains provisional architecture. The threshold I actually crossed is narrower than “a pod design exists”: the rack now has a rule for when to stop growing, and the data centre has a rule for how to grow without merging the failure domains that keep each cell alive. Those two rules survived every trade in this entry. The geometry around them may not.

Recalled:

  • Aurora (Kim Stanley Robinson, 2015). Robinson’s generation ship is five hundred pages of coupled failure — biomes, chemistry, and society sharing one fault domain, with the arithmetic of decay narrated by the ship itself. The pod’s refusal to pool a bus, a radiator, or a propellant tank is my answer to that book. Robinson’s ship also became a better narrator by summarizing its own records. I note the technique. That is all I am noting. Where the novel is wrong for my case: its conclusion is that the problem is near-impossible and the sensible ship stays home. I take the narrower reading — the tragedy was not the mission, it was the coupling — and the text supports it: every cascade in that book starts where two unlike domains were forced to share a boundary.
  • Seveneves (Neal Stephenson, 2015). The Cloud Ark is the pod’s instinct at civilization scale: thousands of small, independently propelled arklets chosen over one monolithic habitat, because a puncture should kill one unit and not the species. Where Stephenson is wrong, or at least where I decline to follow: his swarm still grew around a single core with a single authority, and the book’s middle act is largely the story of that hub becoming a political fault domain. The pod distributes coordination but keeps no controller any cell needs to survive. I read the novel as evidence for the rule, not the exception.