-- Normal-distribution sampling with `rndn mu sigma`.
-- Box-Muller transform under the hood — each call returns one f64 sample
-- from N(mu, sigma). Useful for Monte-Carlo, risk modelling, ML-prep noise.
-- Sample 100 times from N(0, 1); return whether the empirical mean is
-- within 5 standard errors of zero (tolerance 0.5). Stochastic but reliable.
mc-mean-ok>b;s=0;@i 0..100{x=rndn 0 1;s=+s x};m=/s 100;a=abs m;<a 0.5
-- Sample once from N(10, 2). 99.99% of samples lie within mu +/- 4 sigma,
-- so checking the value falls in [2, 18] is safe.
one-shot>b;x=rndn 10 2;r=?>x 2 1 0;u=?<x 18 1 0;t=*r u;=t 1
-- run: mc-mean-ok
-- out: true
-- run: one-shot
-- out: true