Every story in engineering begins with a detail that looks too small to carry the whole piece. This one began there too, with a question that became more useful the longer we left it unanswered.
Start with the smallest compiler pass
The obvious approach was faster, but it flattened the parts that made the subject specific. We kept the rough edges, followed the evidence, and let the structure emerge from what the material actually needed.
The useful abstraction is usually the one you can still explain at 3am.
The result is not a universal method. It is a record of one careful attempt, useful because the constraints remain visible and the reader can decide what belongs in their own work.