A good hash function is a shredder that’s also a fingerprint: change a SINGLE bit of the input and about HALF the output bits flip, unpredictably. That’s the avalanche effect — it’s why a one-character edit to a file gives a totally different checksum, why you can’t nudge your way to a collision, and why the hash reveals nothing about the input. Slide to flip an input bit and watch the output scramble.
The avalanche criterion asks that flipping any one input bit flips each output bit with probability ~½, so a 1-bit input change alters about half the output bits (strict avalanche). This makes a cryptographic hash behave like a random function: it is deterministic (same input → same digest) yet the digest is sensitive, unpredictable, and one-way — you cannot invert it or find collisions by local search. It underpins checksums, digital signatures, and [[the-merkle-tree]]. A fail-loud self-check throws unless a 1-bit input change flips roughly half (10–22 of 32) output bits. ◆ real cryptographic hashing, node-verified.
Illustrated with a small mixing hash (FNV-style), not a full SHA-256; the AVALANCHE property (deterministic, ~half the bits change, unpredictable) is exact for this function and is the defining behaviour real hashes are engineered to meet.