{
  "version": "1.0",
  "type": "link",
  "title": "THESIS 06 // EMPIRICAL PAIRED EVALS & REGIME DRIFT: Multi-Vendor Diversification (Anthropic · OpenAI · Google) & Rebalancing (Australia & Oceania) — Agentic Portfolio Thesis | MARKOWITZ.SYSTEMS",
  "description": "THESIS 06 // EMPIRICAL PAIRED EVALS & REGIME DRIFT (Multi-Vendor Diversification (Anthropic · OpenAI · Google) & Rebalancing) for Australia & Oceania CFOs & quantitative specialists: ρ̂_{AB} = (p_11 − p_A p_B) / √[p_A(1−p_A)p_B(1−p_B)],   (μ_t, Σ_t) ↦ (μ_{t+1}, Σ_{t+1}). Routing and cross-verifying across Anthropic, OpenAI, and Google measurably lowers pairwise error correlation ρ — a falsifiable claim tested via item-level paired evals. Because every model checkpoint release shifts the regime, Σ_t must be re-estimated and agent weights w* rebalanced on every release.",
  "author_name": "MARKOWITZ.SYSTEMS Quantitative Finance & Continuing Professional Education Collective",
  "author_url": "https://markowitz.systems/education",
  "provider_name": "MARKOWITZ.SYSTEMS — Fresh AI Preprints & Professional Upskilling",
  "provider_url": "https://markowitz.systems",
  "cache_age": 86400,
  "thumbnail_url": "https://markowitz.systems/og-card.png",
  "thumbnail_width": 1200,
  "thumbnail_height": 630,
  "url": "https://markowitz.systems/oceania/thesis/th-06",
  "fresh_preprints_url": "https://markowitz.systems/preprints",
  "continuing_education_url": "https://markowitz.systems/education",
  "meta_education_url": "https://markowitz.systems/meta-education"
}