SHA-256 & the avalanche effect

Whatever you type is first converted to UTF-8 bytes — a real difference from "characters": an emoji or accented letter can take up several bytes, and SHA-256 hashes those bytes, not an abstract idea of "characters".

SHA-256 only processes complete 512-bit (64-byte) blocks. Your message is padded: a single 1 bit is appended, then as many 0 bits as needed, then the original bit-length as a 64-bit number — leaving a result that's an exact multiple of 512 bits. For a short string like this tool's defaults, that's a single block.

That padded block runs through 64 rounds of bitwise rotation, shifting, and modular addition, updating eight 32-bit working variables derived from SHA-256's fixed initial constants. The result becomes the chaining value fed to the next block — the Merkle-Damgård construction covered in the Hash functions module — or, since there's only one block here, the digest itself.

Each of those 64 rounds is designed to spread every input bit's influence across the entire internal state within a handful of rounds, the same avalanche principle behind AES. Change a single UTF-8 byte at the very start, and by round 64 essentially every one of the 256 output bits has an even chance of flipping — exactly what you see comparing the two digests above.