FWIW...
..never found a use for it myself, but if you ever need to calculate the side 's' of a square having equal area to a circle of radius 'r' you would normally do...
s= (pi * (r**2))**0.5
... which is a relatively heavy calculation if you have to do it a lot of times.
Instead, you can reduce this overhead quite a bit by first precalculating an alternative constant 'k' to use instead...
k= (pi**0.5) - 1
...which works out to 0.7724538509...
The side 's' of the square can now be calculated by...
s= r + (r * k)
...which is a fair bit easier to do.
And while we're talking about Pi...
To keep my daftness muscle in trim I once wrote some distributed(*) code to calculate Pi by using random numbers. Generate a pair of random numbers between 0.0 & 1.0, treat them as x/y coordinate vectors and sum them: if the sum of the two vectors is > 1.0 they fall outside a quadrant of circle radius 1.0 whereas if they sum to <= 1.0 they're inside the quadrant of circle.
Do this a lot of times and keep track of the total number of pairs (t) and the the number of pairs that fall inside the circle (i)
Pi = (i / t) * 4
It's probably the least efficient way of calculating Pi but it works.
(*) distributed because you really need to do trillions of pairs to get any accuracy.