The original note, as written at the time
Start with a conventional docking tubular type space station equipped with robot arms. Next, we start launching a series of large thick omnipanels that we can build up gradually into a large firm cube.
Ongoing development
Start with a conventional docking tubular space station equipped with robot arms. Then launch a series of large, thick omnipanels, to be built up gradually into a large firm cube.
The method is the idea, and it is a good one: a station that can grow in panels rather than in modules. Modules require a launch vehicle sized for the whole thing; panels require only a launch vehicle sized for one face, and the station grows by adding faces, each one carried by the last.
The robot arms are the enabling detail, because a panel needs to be held while it is attached and inspected while it is attached. That is exactly what a docking station’s arms already do.
The cube has a property that a cylinder does not: its volume grows as the cube of the face length, so every doubling of the structure multiplies the interior by eight. That is why the shape is worth the awkwardness — a cube is the cheapest way to buy volume once you can build in panels.
The engineering problems are radiation and thermal, and the cube makes both worse, because a large flat face is a large flat target and a large flat radiator. The panels have to be thick for shielding, which the note already says, and the corners have to be designed for the concentration of stress.
The honest comparison is a building. This is a construction project conducted in orbit, with robot arms as the labour, and it is much closer to how we build on earth than how we have built in space.
Build out, not up.