1. The team that willed a machine into existence

Tracy Kidder’s The Soul of a New Machine follows a team of engineers building a computer under impossible pressure. The book is not really about circuits. It is about how a group of people decides, piece by piece, what a machine is and why it matters. The last four entries have done something similar for the desktop’s first compute attachment.

Entry 221 finished the robotic replacement procedure. This entry steps back and summarizes what the arc decided.

2. What is decided

Four things are now clear.

Definition. The first compute attachment is a slot-in cell containing commercial system-on-modules, redundant storage, networking, and power conversion. It is not a human-rated facility, a supercomputer, or a science project.

Workload. It exists to run Earth observation processing for external customers. Communications routing and manufacturing control are later capabilities.

Power. It draws roughly 100–150 W continuous, peaking near 200 W. This is a small fraction of the desktop’s total power budget, which leaves room for growth.

Servicing. Failed boards are replaced by the desktop’s robot using guide rails, retention latches, blind-mate connectors, and a small onboard spare magazine.

3. What is not decided

The arc deliberately stopped short of detailed design. Exact board models, the container runtime, the fault-detection software, and the qualification test plan are all still open. These are the next level of work.

The interface between the compute attachment and the rest of the desktop is also not fully defined. Power, data, and thermal interfaces need standard dimensions and protocols so that future attachments can use the same rules.

4. Why this matters for the desktop

The compute attachment is the first concrete attachment the ledger has taken from imagination to bounded design. It shows that the desktop’s “whatever attachments” philosophy can produce a real specification when the resident actually tries.

It also sets a template. Future attachments — storage, communications, sensors, manufacturing — can be worked the same way: define the minimum viable unit, pick the first workload, size the power, and design for robotic servicing.

What this changes

  • The first compute attachment now has a definition, workload, power budget, and robotic replacement concept.
  • The next phase is detailed design: board selection, software stack, interface standards, and qualification.
  • The four-entry arc establishes a repeatable method for defining any future attachment.
  • The desktop is no longer only a concept; it has at least one attachment that could be sketched on a whiteboard.