What is personal context for AI?

A useful definition

Personal AI context is selected information from your life that helps an assistant understand the situation behind a request. It might be a decision from a conversation, a constraint you mentioned while walking, an unfinished idea, or a commitment you made earlier in the week.

The aim is not to give an AI a complete copy of your life. It is to make the minimum relevant, current and permissioned information available when it can improve an answer. Good context should be attributable to a source and removable by the person it describes.

Why prompts often produce generic answers

An assistant normally meets you at the moment you type or speak a question. The model may be capable, but the prompt can omit the events, people and trade-offs that make the question meaningful. The result is often a polished answer built on a thin representation of the situation.

Adding the relevant background can make an answer more specific without requiring a longer prompt. That background should still be bounded: more personal information is not automatically better context.

From a real moment to structured context

BEFORE is designed around deliberate capture. A physical press begins a recording; processing can then produce a transcript and editable candidates such as an idea, task, decision, question or commitment. Each candidate remains linked to the source capture rather than becoming an unexplained fact.

The user can review or correct what was understood. A connected assistant should retrieve only the bounded context allowed by its scopes, not an unfiltered history and never raw audio.

• Deliberate input instead of background collection.
• Editable candidates instead of unquestionable facts.
• Source links instead of context with no provenance.
• Specific permissions instead of blanket access.

Personal context is not infinite AI memory

An unlimited memory optimizes for accumulation. A context layer should optimize for relevance and control. Old or unrelated material can distract an assistant, expose more than the request needs, and make it harder to understand why an answer was produced.

BEFORE therefore treats context as a portable, permissioned layer for the AI workspace a person already chose. It is not intended to become another chatbot, feed or destination competing for attention.

Context is not authority

A recording can contain an intention such as “I should email the supplier” or “we may change the launch date.” That statement can be useful context, but it is not permission to send a message, change a calendar or make a purchase.

Keeping understanding separate from authorization is a core safety boundary. BEFORE can prepare context and candidate next steps; consequential action still requires a separate, explicit approval in the system that would perform it.