I built mantic-thinking for consistent multi-factor analysis. 14 tools (7 friction, 7 emergence), immutable kernel: M = (sum(W * L * I)) * f(t) / k_n.
Ran 5 tests. Key findings:
The "critical window" pattern
Friction M > 0.7 AND emergence M > 0.7 = unstable equilibrium. High risk + high alignment for action. Neither tool alone catches this. Tested on Tesla: regime c
onflict (0.715) + confluence window (0.810) → "small position, tight stops."
Cross-domain transfer works
86% pattern consistency across healthcare, finance, cyber, climate, legal, military, social. Same mental model, different domain weights.
Temporal kernels enable timing
8 kernels model evolution. Exponential decay = "monitor." S-curve = "act at inflection." Makes "when to act" explicit.
LLM integration is load-bearing
Framework needs LLM front (NL → parameters) and back (M-scores → recommendations). Tested cyber attribution: extracted params, calculated, synthesized "attribut
ion gap—isolate server, 48-hour window."
Architecture holds up
Zero-variation core kernel. Graceful NaN handling. 0 crashes. Boundary conditions handled.
I built mantic-thinking for consistent multi-factor analysis. 14 tools (7 friction, 7 emergence), immutable kernel: M = (sum(W * L * I)) * f(t) / k_n.