C.17:9.2 - Software and algorithmic design

Software design. In this worked example, ETL-Parallel-12 is compared with 40 admitted internal pipeline designs through ASTGraphSimilarity-3, whose declared Novelty construction uses calibrated [0,1] similarities. The Method compares AST-graph representations produced by the same declared parser and projection for the evaluated design and every corpus design; the result states that runtime configuration and deployment topology are not preserved by that projection. Its Novelty result is 0.36. A fixed-workload benchmark reports an 18% lower p95 latency than the serial baseline, but the segregation-of-duties test fails. The benchmark run, corpus edition, nearest-neighbour report, and policy test are the evidence. The result therefore supports a latency gain but leaves the design ineligible for the stated use; redesign the isolation boundary and repeat the affected tests before any pool or choice decision.

Algorithmic search Work. SpikeSet-7 contains nine dated attempts in three declared approach classes over six hours. Tagged Work records support DiversityOfSearch = 3 classes; the first runnable output appeared after 2 h 10 m. The held-out viability test was never run, so Time-to-First-Viable is not established. The practical action is to run that test, not to relabel time-to-first-runnable as viability.