SAI Methodology Changelog
Format: version | date | change | rationale. New versions ship at the top. The currently-published page is rendered from sai_methodology_kardashev.md + Stage 1 audit findings + live K_access data by src/pipeline/sai_methodology_writer.py.
v1.0 — 2026-06-10 — Initial publication
Change: First public methodology page with the Kardashev / Sagan / Barrow / Zubrin lineage, the three-axis (Outward Capacity / Inward Sophistication / Economic Depth) presentation of the 17 composite sub-metrics, the K_access companion stat (log10(M) / 9) with live trailing-12-month value and 1957- present annual series, and the §6 ARI division-of-labor framing.
Rationale: The SAI TRANSPARENCY anchor on /sai/ was an empty JS-hydrated container; ARI was referenced publicly but never defined. Closing both gaps in one workstream aligns the methodology page with the way the composite is actually computed (17 factors, not 6) and ships K_access as a judgment-free physical-derivative cross-check that any reader can reproduce.
SAI composite math: Unchanged (frozen). The three-axis grouping in §3 is presentational only. Composite stays at weighted_sum / total_weight per src/sai/index.py:calculate(). Verified byte-identical pre/post: composite_score=60.1, maturity_label=Mature, 17 live factors.
K_access:
- Formula:
K_access = log10(M) / 9, M in kg. - Anchors: M=1 kg/yr → K=0.00 (pre-spaceflight floor);
M=10⁹ kg/yr (1 Mt — interplanetary logistics threshold) → K=1.00.
- Current method: per-launch mass uses vehicle-registry LEO throw-weight as
consistent capacity-proxy upper bound. Method flag per year: mixed (≥95% coverage), estimated (50-94%), insufficient (<50% or no rows). Live current at publish: 3,825 t/yr → K=0.73.
- v1.1 roadmap: dual-source per-mission measured mass for 2010+ records.
Sustainability stance (§7): v1.0 composite does NOT include a debris/ sustainability adjustment. Catalog tracks decay/debris but no penalty term exists in src/sai/weights.json. Documented as the first item on the methodology roadmap; no implied shipped feature.
ARI teaser (§6): Stays within the Stage 1 Agent 4 allowlist — factor count (11), category structure (Operational 30% / Financial 40% / Market 30%), confidence-tag language (reported/derived/estimated), and TRL × IRL framing. Per-factor names, formulas, weights, and rationale strings remain behind the Intelligence-tier gate at /intelligence/#ari-drill-in.