Autocorrect software was originally developed by Microsoft engineer Dean Hachamovitch in the 1990s as a helpful glossary feature to instantly fix common spelling mistakes in Word.

A tiny replacement list, built into the act of typing

In Word, a typo such as “teh” can be corrected before it becomes part of the sentence the writer is composing. The key is that AutoCorrect does not need to understand the sentence’s meaning: it can match a familiar mistyped sequence to a stored replacement, turning “teh” into “the.” That makes the feature more like a glossary of known errors than a spell-checker weighing the whole passage.

Why make the correction automatic?

A spell-checker typically flags a possible problem and leaves the decision to the writer. AutoCorrect takes the extra step of making a replacement immediately. For mistakes that are predictable and unambiguous, that saves a pause: the writer can keep going instead of stopping to inspect an underline or choose from suggestions.

That convenience depends on a limit. A word sequence alone cannot always reveal what someone meant. A string that is usually a typo could be intentional in a name or specialized term. Automatic correction works best when the replacement is highly likely to be right—and when the writer can notice and reverse a mistake.

A small intervention with a lasting role

Hachamovitch’s 1990s feature brought correction into the typing process itself, rather than treating it only as a review task after writing. Its simple logic—recognize a common error, substitute a likely intended form—also captures the broader bargain behind autocorrect: less effort for routine slips, in exchange for trusting software to decide when a correction is safe.