Two physicians look at the same chest X-ray. One reaches a diagnosis in three ops — a glance, a pattern match, a conclusion. The other requires twenty-eight ops — systematic review of each lung field, comparison against mental templates, elimination of differential candidates, tentative conclusion, recheck. Both arrive at the same diagnosis. The X-ray is the same. The goal is the same. The context is the same. The cost is different by nearly an order of magnitude. Now hold the processor constant and vary the task. The experienced physician processes a classic pneumonia presentation at three ops and an unusual autoimmune presentation at twenty-two ops. Same physician, same goal — diagnose the patient — but the cost landscape across different presentations defines a profile. Some tasks are dissolved to zero. Some require full pipeline engagement. Some are outside the physician's domain entirely. The shape of that cost landscape is this physician's expertise, made visible. These observations raise a precise mathematical question. The cost of processing — the op count a specific processor requires for a specific task — is a measurable quantity. When you collect these costs across a set of tasks for one processor, you get a profile. When you collect profiles across multiple processors, you get a space. Does that space have geometric structure? Can you define a meaningful distance between two profiles such that the distance satisfies the axioms of a metric space? If so, expertise has a geometry. Skill gaps are distances with direction. Learning is a trajectory through a space. Clusters of similar practitioners are formally identifiable objects rather than informal impressions. The question of how far apart two processors are — in skill, in specialization, in development — becomes a question with a numerical answer. The vocabulary is small and builds in order. Processing is what any system does when it must act on information — a CPU scheduling processes, a surgeon operating, a pilot navigating, a developer debugging. The unit of processing is the **op**: one irreducible transformation by one processor. A diagnostic question, a mirror glance, a cache lookup, a line of code read and understood — each is one op. Processing entropy is the op count a specific processor requires for a specific task. It is receiver-dependent: the same chest X-ray has processing entropy of three at the experienced physician and twenty-eight at the junior resident. When processing entropy reaches zero, the task is **dissolved** — the processor handles it structurally, without consuming its scarce sequential pipeline. The processor's total capacity is bounded by one inequality: total ops multiplied by average op duration must not exceed the available time budget. Dissolved tasks don't count against the budget. That is what makes expertise powerful — the expert's budget is free because most routine processing has dissolved. This paper takes processing entropy as its primitive and investigates the mathematical structure that emerges when you consider many processors and many tasks simultaneously.
Geoffrey Howland (Mon,) studied this question.