Persistent context
Persistent context
Persistent context is information an AI system keeps about a particular person between sessions, allowing later conversations to build on earlier ones. It separates a system that knows you from one merely briefed on you, and provides the material every relational system relies on.
Also called durable personal context, AI long-term memory, AI memory
People often call persistent context a memory feature, which understates it. Choosing what a system keeps about someone also chooses who that person is from the system’s point of view. A record that keeps each stated preference forever fixes a person in place. One limited to the latest exchanges cannot hold a thread.
Storage is the easy part. Selection, revision, and proportion are harder: what deserves retention, what should fade, what needs revision as a person changes, and what should never enter the record? Perfect recall of an offhand remark from two years ago can make a system stranger and less useful than one that forgot it.
Persistent context creates a legibility problem too. People cannot correct a record they cannot see. An invisible memory system leaves someone living beside a version of themselves they cannot inspect or amend.
Why it matters
Persistent context turns a helpful system into a permanent record. Too little leaves the system amnesiac and useless. Too much can build a dossier the person never agreed to and cannot inspect.
Where TAICU stands
TAICU holds that memory must support continuity while staying legible, correctable, and proportionate. The lab treats those as design requirements from the start; a memory system without them resists later repair.
Common questions
- Which facts should an AI retain about someone?
- It should retain enough to preserve continuity: the material that makes later conversations coherent. Proportion is the useful test. Retention should be warranted by its use, and a person should be able to view and correct it on request.
- How does persistent context differ from a long context window?
- A context window is the amount of text a model considers at once. Persistent context is deliberately retained between sessions. Even a huge context window leaves a system forgetful when nothing carries into later conversations.