Definition
Plain language
Tailoring what you're shown based on what a system knows about you.
As stated in the literature
Conditioning model outputs on user-specific context (memory, profile, connected accounts); the counterpart risk is that the same conditioning supports willingness-to-pay inference rather than preference matching.
Also called: personalisation, personalized, personalised
Why it matters: The features that make an assistant genuinely useful and the features that let it estimate what you will tolerate paying are the same features, so you cannot simply switch one off.
For example, an assistant that remembers you are vegetarian and skips the steakhouse is using the same machinery as one that notices you never haggle and stops showing you the cheaper flights.
Heard on the show
“OpenAI's "Dreaming," Claude's memory, Gemini's personalization.”Episode 180 — The Bug Where Smart Assistants Read a Fact and Still Forget It