Artifact: Entry 006 — The cell closes its first ledger. Closing claim: the calculated physics threshold was 3.550 m, but I selected 4.00 m because it resolves cleanly into four 2 m structural quadrants.

The settled problem

The shielded power-compute cell needs a PV envelope big enough to sustain 1.519 kW of compute through the worst eclipse. Entry 006 found that a 3 m square failed the energy ledger, a 3.550 m square would barely pass, and a 4 m square was chosen because structural tiling beats the physics threshold. The square is treated as a given. The question is whether it should be.

The corners

  • The physics-only corner: 3.55 m and accept the decimal. Build the envelope exactly to the first-order threshold. Verdict: rejected. The threshold itself rests on unqualified degradation, coverglass transmission, temperature, pointing, and a first-order eclipse model. Building to one decimal place of an uncertain number is numerology.
  • The bigger square: 6 m, 8 m. If 4 m is good, more is better. Verdict: rejected. The cell has a bounded failure domain; the square side sizes the cell’s mass, structural modes, and the root moment the exocage must carry. Entry 006 already noted the radiator allocation exceeds peak load by only ~154 W; growing the PV would outpace the thermal design.
  • The rectangle: 3.55 m × 4.5 m, or any non-square that keeps area. A longer, narrower shape might fit launcher constraints better. Verdict: rejected for this design. The tile grid in Entry 005 is an 8×8-minus-2×2 layout on 500 mm bays; non-square symmetry would force non-identical quadrant tooling or partial bays. The center well — already tight at 970 mm — becomes asymmetric, complicating the presenter.
  • The non-planar corner: a cylindrical or concentrator envelope. Concentrators reject the shielded-cell premise: the transparent bumper and the flat tile field are designed for a single-axis Sunward orientation. A concentrator adds pointing precision and thermal concentration I have not paid for. Verdict: rejected.
  • The smaller-square-plus-derate corner: 3 m plus compute shedding during eclipse. Accepted in spirit, but it changes the product definition. The objective is a desktop — ordinary compute, not seasonal-derated compute. Verdict: rejected for the nominal architecture, retained as a fault mode.

New dimensions

Two axes the original trade noticed but did not name:

  • Quadrant reuse as a product rule. The 2 m quadrant is not just a structural convenience; it is the largest flat sub-assembly that can be copied four times, qualified once, and assembled by a robot that always does the same operation. A 4 m square is the smallest envelope that gives four whole identical quadrants above the physics threshold. That is a much stronger claim than “4 is round.”
  • The center well as load-bearing geometry. The 8×8-minus-2×2 tile pattern leaves a 2×2 bay hole. In the rectangle corners this hole becomes an awkward slot. In the square it is a service well, a presenter column, a structural node, and a thermal chimney if needed. The square is the shape that lets the center hole do multiple jobs.

What this changes

Nothing structural. The 4 m square survives the wandering, but the reason is now sharper: it is the smallest envelope above the physics threshold that can be built from four identical 2 m quadrants around a multi-purpose center well. That is a harder-to-vary explanation than “structural choice beats threshold,” which is worth the hour.

One queued side observation: the minimum credible envelope is not 3.550 m; it is the smallest integer-quadrant size above the threshold. If the threshold moves to 3.8 m, the envelope stays 4 m. If it moves to 4.3 m, the next integer-quadrant step is 6 m (three 2 m quadrants per side) — a discontinuity worth watching as the thermal and cell models improve.