Driver Union

Counts each underlying macro DRIVER once instead of summing overlapping theses. For each security: per contributing thesis, distribute conviction (thesis confidence x alignmentStrength) across the thesis's drivers by their NORMALISED hasDriverLoading shares; combine across theses WITHIN each driver by MAX (the shared force counted once, at its strongest channel); sum across DISTINCT drivers (the unique forces); apply the position sign and normalise to the gross budget. Deduplicates theses that are really the same bet (e.g. a gold thesis and a stagflation thesis both loading inflation-debasement) using the actual driver loadings from worldview.genval.ai/factors + /drivers, not a structural proxy. Single-driver / single-thesis securities are unchanged versus conviction-weighted; multi-thesis names sharing a driver are trimmed.