Counterfactual Loom v1: Universal Collapse Knot Families Across Finance, Crypto, and Seismic Systems

This record contains the first experimental implementation of the CNT Counterfactual Loom (CFL), a Cognitive Nexus Theory (CNT) module for quantifying “near-miss futures” around real-world events. In CFL, each historical window of a system (finance, crypto, or seismic) is treated as a counterfactual fiber: a bundle of outcomes from a fixed family of simple policies applied to the same data. For each window, we compute: • fiber_std_reward — spread of rewards across policies (policy disagreement)• fiber_collapse_weight — fraction of policies with catastrophic loss (reward < −1.0)• fiber_redeemability — fraction of policies with strong positive reward (reward ≥ +0.5) Using this shared policy bundle and reward model, we apply CFL to four domains: BTC-USD (crypto), SPY (equity index), VIX (volatility index), and a large set of M7.9–M9.0 seismic traces (Sumatra 2004, Tohoku 2011, Maule 2010, etc.). This enables direct comparison of fiber geometry between markets and earthquakes. We show that: The distribution of fiber variance and collapse weight exhibits a common discrete structure across domains when grouped by the number of catastrophic policies k (0–4 of 6). Each knot family Kₖ has an essentially universal geometry (e.g. K₂ and K₃ knots have high variance and collapse weight, with zero redeemability in all domains). The frequency of these knot families differs strongly by domain. Finance and crypto visit catastrophic knot families (K₂, K₃, K₄) but spend most of their time in K₀/K₁ (no or single-policy collapses), whereas seismic traces exhibit a much higher density of K₂/K₃ knots, indicating extended regions where multiple policies fail together and no policy is clearly safe. Event-level analysis of the M9.0 Sumatra 2004 seismic trace shows a broad band of high-susceptibility windows in which many simple policies are simultaneously wrong, consistent with CNT’s view of disasters as “dense counterfactual knots” in policy space. The release bundle includes:• Domain geometry and tail statistics (cfl_domain_geometry_summary_v1.csv, cfl_domain_tail_geometry_summary_v1.csv)• Knot family summaries for all domains (cfl_knot_family_summary_v1.csv, cfl_knot_family_highsus_summary_v1.csv)• High-susceptibility window sets and annotations for finance/crypto and seismic (cfl_high_susceptibility_nodes_*.csv, cfl_seismic_high_sus_nodes_v1.csv)• Event-level Loom tables and plots for Sumatra 2004 (cfl_event_M9.0_20041226_*.csv/.png)• Full fiber summaries (with and without seismic) and a Markdown overview (CNT_CFL_v1_SUMMARY.md) This record is designed as a reproducible CFL v1 module within Cognitive Nexus Theory: it defines the basic counterfactual fiber metrics, provides cross-domain evidence for universal collapse knot families, and establishes a template for future CFL extensions (more complex policies, richer reward models, and additional domains).