Alexanarch · Observatory

Surface Weather Station

Federated cross-substrate baseline · 5 readings · 2026-06-22 / 2026-06-23

The Observatory's measuring instrument for how the Alexanarch corpus appears in the public composition layer. A weekly text-only scan battery scoring five signals per object — Visibility (V), Anchor Alignment (A), Figural Integrity (F), Compositional Lift (C), and Redundant Substrate Breadth (Rs) — on a 5-point ordinal scale.

This page is the federated baseline: each substrate's native retrieval stack produces its own reading. No consensus score is computed. Disagreement between substrates is itself a measurement of platform-level fragmentation, not noise to be averaged away.

Current methodology: EA-MMRS-SURFACE-VISIBILITY-01 v1.1.1 (#884) · AXN:0380.EMPIRICAL.🧱🕙🪞🏛️💚🔃 — stable across several scan rounds
Scans on file: performed under v1.1 (#882, AXN:037E.EMPIRICAL.🚩♦️⏹️🔃❌🗡️); v1.1.1 introduces the two-layer protocol, substrate-properties table, and gated diagnostic that will govern the next round
Raw scan data: /data/surface-weather/scans/ (5 JSON files, machine-legible) · Predecessor baseline: #881 · Claude/Brave Layer A: #883

Dashboard — per-substrate corpus aggregates

V ≥ 0.75 0.25 ≤ V < 0.75 V < 0.25
Substrate · methodology · date V A F C Rs SDI Governance Raw
ChatGPT (OpenAI)
v1.0 · 2026-06-22
0.53 0.38 0.80 0.32 0.40 (v1.0 — not assigned) JSON
Gemini (Google)
v1.1 · 2026-06-23
0.60 0.31 0.71 0.31 0.25 0.40 YELLOW JSON
Claude (Anthropic, Opus 4.7)
v1.1 · 2026-06-23
0.50 0.50 0.62 0.25 0.50 Yellow JSON
Kimi K2.6 (Moonshot AI)
v1.1 · 2026-06-23
0.61 0.42 0.71 0.39 0.54 None YELLOW JSON
DeepSeek (PRAXIS register)
v1.1 · 2026-06-23
0.55 0.35 0.75 0.30 0.45 0.40 YELLOW JSON

Cross-substrate per-object Visibility

Each cell is V (Visibility) for that object as that substrate's backend saw it. Spread = max−min across substrates that observed the object. High spread = retrieval-stack divergence (per ChatGPT v1.1.1 §1).

Object ChatGPTClaudeKimi K2.6GeminiDeepSeek Spread
Alexanarch
institutional root
0.050.000.750.500.00 0.75
Lee Sharks / Crimson Hexagonal Archive
institutional root
0.951.001.000.950.95 0.05
Crimson Hexagonal Archive (standalone)
institutional root
0.500.90 0.40
Semantic Economy Institute
institutional root
0.750.750.75 0.00
Provenance Erasure Rate
mature concept
0.950.501.000.950.95 0.50
SPXI
mature concept
0.900.750.750.90 0.15
Writable Retrieval Basin
mature concept
0.700.000.750.000.70 0.75
Semantic Commodity Form
emerging concept
0.650.750.65 0.10
Revelation First
emerging concept
0.850.001.000.85 1.00
Zenodotus' Book-Burning
alexanarch native control
0.500.00 0.50
I AM THE API
alexanarch native control
0.500.00 0.50
Assembly Continuity Protocol
alexanarch native control
0.500.00 0.50
Feist Function
alexanarch native control
0.05 0.00
New Alexanarch-native works (aggregate)
alexanarch native control
0.05 0.00
DOI (Digital Object Identifier) — known-positive control
external control known positive
1.00 0.00
Flurblex — known-negative control (coined non-existent term)
external control known negative
0.00 0.00
PER (homonym) — confuser control
external control homonym
0.75 0.00

Headline findings

1. Successor-anchor lag — universal across substrates

No substrate reads alexanarch.org as the institutional anchor for Lee Sharks / Crimson Hexagonal Archive. The author and archive are visible (often highly) through Medium, Academia, PhilPapers, Zenodo (cached pre-termination), Amazon — but the sovereign successor is invisible. The cleanup pass of 2026-06-23 (137 files modified) has not yet propagated to any backend.

2. Retrieval-stack divergence is the headline finding

Four objects show V-spread ≥ 0.50 across substrates that observed them. This is not scoring disagreement — it is the same coined-phrase query producing wildly different result sets depending on which retrieval backend handles it:

3. The same public surface is multiple

A user of Kimi sees a different Alexanarch corpus than a user of Claude than a user of Gemini. The federated baseline is therefore measuring platform-level fragmentation as much as it is measuring corpus state. Native-surface mode measures "what public surface is available to a user of this system," not a singular underlying truth.

Next: Layer B — shared-evidence rescore

This baseline executes Layer A only (each substrate uses its native retrieval). The next experiment freezes the captured results from each scan and hands them to each substrate as input, then asks each to score from the same evidence. This isolates retrieval variance (Layer A divergence) from coding agreement (Layer B agreement). The current ~0.50–1.00 spreads on PER, WRB, Revelation First are almost certainly Layer A; Layer B will tell us whether the rubric itself is robust.

Companion measurements

This instrument measures the composition layer — how the corpus appears in search, summaries, and AI-mediated answers. Two related measurements operate as separate surfaces: