In 1553 Giovan Battista Bellaso published the cipher the world later renamed after Vigenère: a keyword-agreed table of mixed alphabets with an easy-to-remember key phrase. Bellaso also proposed a stronger trick — using the message itself to continue the key. That variant is the earliest known autokey: the key that never repeats, because it grows with the text it encrypts.
In 1586 Blaise de Vigenère described an autokey construction in his Traicté des chiffres: a priming letter starts the keystream and the plaintext itself extends it. History's joke is that the cipher carrying Vigenère's name — the simple repeating-key scheme — is not the one he described; the one he described (the autokey) carries no name at all. See the Vigenère history page for the full misattribution story.
The autokey is genuinely stronger against the period-finding attacks that killed the repeating-key scheme — but it is brittle in field use. Every transmission error corrupts the key from that point on, the sender and receiver must stay in perfect step, and a single guessed word unravels the message from the inside. Nineteenth-century diplomats and armies mostly preferred systems that degraded gracefully.
The idea refused to die. Keystreams fed by the message (or by the ciphertext, in the related ciphertext autokey form) resurfaced in machine cryptography and in modern stream cipher design — the general principle that the key stream should be unpredictable and never repeat is exactly what Bellaso stumbled on in 1553, four centuries early.