Most meeting tools treat each chat thread like a blank slate. Scriba doesn't. Every meeting and project carries a running memory — a single blob the AI rewrites after every assistant reply and you can edit directly.
Why persistent memory matters
Decisions drift across calls. Someone asks in week three what you already agreed in week one. Without memory, you re-explain context or grep old transcripts manually. Brain memory rolls forward so new threads inherit what mattered.
- Injected into every meeting and project chat system prompt.
- Consolidated automatically after each assistant reply.
- Editable in the Context tab when you need to correct or pin a fact.
- Synced to AWS with last-write-wins semantics.
How consolidation works
After each reply, a background job distills the exchange into the existing memory blob. It's fire-and-forget — it never blocks the visible response. Markers like {{ref:…}} and {{contact:…}} are preserved verbatim so chips still resolve in the UI.
Memory is not a summary of the last message. It's the accumulated state of the meeting — decisions, owners, open questions.
Rolling up into projects
Linked meetings contribute their brains to project chat. Instead of per-meeting highlight snippets, you get one project-level memory that reflects the arc of the work. Cross-meeting questions stop feeling like archaeology.