🎯 Why is this curve everywhere? Two superpowers: the Central Limit Theorem and the fact that Gaussians stay Gaussian when you combine them.

Central Limit Theorem (CLT): take almost any distribution — lumpy, skewed, weird — draw a bunch of values, and average them. Do that many times and the averages pile up into a bell curve. It doesn't matter what you started with. This is why so many real-world measurements are Gaussian: they're secretly sums of many small, independent effects.

In neuroscience that's the norm: ask a subject to point at a sound and their responses are Gaussian; measure fMRI activity and it's Gaussian. Countless tiny influences add up, and the CLT does the rest.

🔗 Gaussians are self-multiplying. Multiply two Gaussians together and you get… another Gaussian. Its new mean is just a weighted average of the two old means (weighted by their precisions):

N(μ₁, σ₁²) · N(μ₂, σ₂²) → N(μ_new, σ_new²)

The product of two bells is a third bell, with a linearly-combined mean. This closure property is what makes the Gaussian math tractable.

This is a bigger deal than it looks. That "stays in the family" property is why Gaussians power filtering, linear systems, and optimal control — combine your prior belief (one bell) with new evidence (another bell) and the answer is a clean, closed-form bell. No messy integrals.

The two knobs also line up perfectly with the statistical moments you already care about: the mean μ is the first moment (the average / expected value), and the variance σ² is the second moment (the spread). And it all extends to many dimensions — μ becomes a vector, σ² becomes a covariance matrix Σ — which is where this meets linear algebra.

🤖 AI tie-in: neural-network weights are typically initialized from a Gaussian; the noise in diffusion image-generators is Gaussian; and "assume Gaussian errors" is the hidden assumption behind ordinary least-squares regression — the workhorse you'll meet in Computation Caves.