nx_growth_series.nx

nx_growth_series.nx -- THE TIME SPINE for the exec/managerial/operational wedding cake. nx_pm_board emits the three tiers correctly. What the estate had NOWHERE was TIME: measured 2026-08-15, ecomat hist EMPTY, frontier hist EMPTY, comparewatch unseeded, token_efficiency.log 1417B -- so no board could answer "how did this change since last week". A LEVEL CANNOT EXPRESS GROWTH: growth is a DERIVATIVE and needs two samples in time. This organ adds ONLY that dimension. Every number it records is COMPOSED from nx_pm_board through tr_run_capture and every historical number is READ from journals already on disk. Zero measurement logic is duplicated: one ruler per figure. nx_growth_series beat [log] [board-elf] -> append ONE dated row; a BEAT-FAIL is ALSO a row nx_growth_series report [log] -> d/w/m deltas + the per-axis HORIZON nx_growth_series mine [journal] -> retro-computed workstream history, FULL population nx_growth_series selftest [scratchlog] -> gate teeth incl. named negative controls ROW SHAPE -- the ECOMAT shape already proven in knowledge/status/ecosystem_maturity.log, so reader idiom transfers. Line-anchored, each field name appears EXACTLY ONCE, canonical verdict LAST so a positional reader can anchor on it. A FIELD THAT COULD NOT BE READ IS -1, NEVER 0. "Could not look" and "measured zero" are different facts, and a series that confuses them reports a collapse that never happened. NO GUESSED CEILINGS. Every buffer here is sized from the thing that fills it -- needles from the key, emit lines from the field count, the mine day-grid from the journal's own epoch span. The one remaining bound (GS_BOARDCAP) guards a SUBPROCESS's output, whose size is genuinely unknowable in advance; it is named for that single purpose and its truncation ANNOUNCES as a RED row rather than being absorbed. A ceiling that has to be guessed is a defect generator in both directions. license_tier: ORIGINAL No hw writes (Rule 26). expect_exit: 0