1. The astronaut who counted potatoes

Andy Weir’s The Martian is full of elegant orbital mechanics, but its real engine is inventory management. Mark Watney survives because he knows exactly how much food, water, power, and air he has, and he defines his problems tightly enough to solve them with what is at hand. The desktop has spent many entries imagining rooms. This entry tries to define one room tightly enough to build it.

Entry 217 stepped back from the wonderings. This entry steps forward and picks the first attachment: compute.

2. Why compute first

The desktop objective says “powered and connected desktop with whatever attachments.” Compute is the attachment that makes the other attachments useful. Without it, the desktop is a power and thermal bus with nothing to do. Sensors need compute to process data. Manufacturing needs compute for control. Communication needs compute for routing. Compute is the first attachment because it unlocks everything else.

3. What a first compute attachment is not

It is not a general-purpose cloud server. It is not a supercomputer. It is not a radiation-hardened mainframe. Those might come later. The first compute attachment is a minimum viable processing unit that proves the desktop can run useful workloads in LEO.

It is also not a science project. It should use available components, accept standard software interfaces, and fail in predictable ways.

4. A minimum viable definition

A first compute attachment could be defined as follows:

  • Form factor: one standard desktop cell slot, with blind-mate power and data connectors.
  • Processing: a cluster of commercial system-on-modules, selected for power efficiency and fault tolerance at the software level rather than custom radiation hardening.
  • Storage: redundant solid-state storage with error correction and periodic scrubbing.
  • Network: redundant Ethernet or SpaceWire links to the desktop backplane and to external radios.
  • Thermal: conduction-cooled through the cell wall to the desktop radiator loop.
  • Software: a container runtime or virtual machine monitor that lets customers deploy workloads without requalifying the hardware.
  • Reliability: single-board failure should not take down the attachment; failed boards should be identifiable and replaceable by the servicing robot.

This is not a final specification. It is a boundary around the problem.

5. The radiation question

LEO radiation is real but not extreme at moderate inclinations and altitudes. Commercial parts can survive if the software watches for errors, retries operations, and reconfigures around failed units. This is radiation tolerance by architecture, not by custom silicon. It is cheaper and faster, and it matches the desktop’s philosophy of using standardized parts.

6. What would make it real

To move from definition to design, three questions need answers:

  • What is the actual power budget per compute slot? This sets the number of boards and their clock speeds.
  • What is the expected workload? Earth observation processing, communications routing, and manufacturing control have very different compute and storage needs.
  • What is the robotic replacement procedure? The attachment must be designed so the servicing robot can pull a failed board without disturbing neighbors.

These answers do not require new science. They require focused engineering work.

What this changes

  • The first desktop attachment to define concretely is compute, because it enables every other attachment.
  • A minimum viable compute attachment can be built from commercial components with software-level fault tolerance.
  • The definition leaves room for future upgrades without requiring a complete redesign.
  • The next entry should pick one of the three open questions — power budget, expected workload, or robotic replacement — and work it toward a decision.