Library / First Principles Framework (FPF) - Core Conceptual Specification
Jump to passage
In this reading

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 15:41:27 UTC · snapshot created 2026-10-03 15:43:46 UTC · last check 2026-10-03 15:55:10 UTC

B.1.6:6 - Archetypal Grounding

Engine test programme. C.16 measurement Work attributes fuel mass, electrical energy, operator time, and emissions values to exact subjects under their Scales, models, calibration bases, windows, and uncertainties. Each has its own result episteme. Exact test-run occurrences and obtaining A.15.1 Work-part relations define the included Work set; independently declared C.27.TA overlap facts state shared timing. A test-cell or engine phase enters only through the carrier’s identity rule and proper phase relation. Shared warm-up energy is recorded under the exact temporal and resource-use facts. Dated aggregation Work applies ProgrammeResourcePolicy-v3, which allocates warm-up energy once and propagates input uncertainty. The B.1.6 result is a typed resource vector plus qualifications; a C.2.1 episteme states it. A later emissions verdict remains separate evaluation Work and result.

Manufacturing cell. Welding and painting are two dated work occurrences. Electricity, gas, consumables, and labor time are separate resource Characteristics with measurement-result epistemes. A shared extraction fan and overlapping operator time require direct shared-use facts and an allocation policy. The resource ledger represents those facts. Establish any separately claimed Work-part relation or frame transformation through its direct pattern.

Model training. Epoch labels alone do not establish work parts. Ground the training work and exact slices, then recover C.16 measurements for compute energy, storage traffic, and operator time. Aggregation work applies an edition-pinned policy to those result epistemes. The training Method, any algorithm text that qualifies as its MethodDescription, and the dated training Work remain distinct; trained-model identity, fairness result, provenance, assurance, and deployment decision stay with their subject patterns.