2025-04-16
Prompted by Show HN: Unsure Calculator – back-of-a-napkin probabilistic calculator | Hacker News, which is the same guy who created that Knights of San Francisco game and related posts about game design theory, he has Unsure Calculator and a more elaborate Unsure app — both are entries in the space of little calculators that provide uncertainty intervals on the numbers. Comments in the HN thread brought up several projects that I was looking at recently, in particular Squiggle and Guesstimate. This person says that Emacs calc mode has both “interval forms” (arithmetic on both ends of an interval) and “error forms” (close to this idea of confidence intervals). This one points out that Qualculate handles this as well, but maybe not with nice visualizations. (Yeah, see the manual: Chapter 6. Propagation of Uncertainty and Interval Arithmetic.) From another comment, nice, a short and dirty implementation in Python with NumPy. Causal: The finance platform for startups is an enterprise-focused app that does this type of modeling.
Mentioned in another comment in the above, a nice paper takes another look at the Drake equations and “Fermi paradox.”* The point is that if you look at the Drake equation as a probability / Bayesian modeling exercise and actually assign distributions to the parameters, you end up with a final distribution of intelligent civilizations that indicates the chances of 0 other ETIs in our galaxy is at least 20% and maybe higher. How do you assign reasonable distributions to the unknown parameters? They do it two ways: first, they randomly sample combinations of values reported in this literature over the past many decades. Second, they just pick log-uniform distributions based on conservative assumptions. Both approaches given broadly similar results. Absolutely love it. A fun project idea is to reproduce this with PyMC in a notebook or similar.
Finally, and not related to the above, here’s The Dusa Programming Language.