IEEE 754. Most of us treat division by zero as a crash. The standard does not. Under IEEE 754, 1/0 = ∞ Infinity is a first-class value, not an exception that aborts the process. NaNs, signed zeros, and gradual underflow all exist so that computation can continue past the point where ordinary arithmetic would have terminated. When a quantity approaches zero, the default is not to stop but to stay — lifting the result into a non-finite domain where the rest of the program remains defined. Recursion and infinite loops are special cases. The best floating-point code is yet to come →∞❤️