Two big upgrades to how your Second Brain understands and protects your memories — plus reliability improvements. It's all automatic and fully backward compatible: deploy it, and your existing memories keep working while new ones get smarter.
🛡️ Your important memories are now protected
Until now, the newest memory always won — so a quick, low-quality note could quietly overwrite or delete a decision you'd confirmed or a preference you'd set. Not anymore.
Second Brain now understands a memory's status:
- Canonical — confirmed, authoritative memories (a decision, a durable fact, a permanent preference). These are recognized automatically as you capture them and are protected from being overwritten or deleted by newer, conflicting notes.
- Draft — tentative or still in progress.
- Deprecated — no longer accurate. These drop out of recall so they stop cluttering your results, but they're kept for your records — nothing is silently destroyed.
How it helps you: you can trust that what you've confirmed stays put. When a new note contradicts a canonical memory, Second Brain keeps the trusted one and files the newcomer as a draft for you to reconcile — instead of clobbering your source of truth. And you can pin or retire any memory yourself whenever you want.
🗂️ Events vs. knowledge — recall what you actually mean
Every memory is now automatically sorted into one of two kinds:
- Episodic — things that happened: events, decisions, milestones ("Shipped v2 on June 13").
- Semantic — things you know: facts, preferences, knowledge ("We use Postgres for all services").
How it helps you: "What happened around the launch?" and "What do we know about auth?" are different questions — and now you can ask them separately. Recall can filter to just events or just knowledge, so you get the right kind of answer instead of a jumble. (This also lays the groundwork for timeline and knowledge-base views coming next.)
🔧 Reliability
- Classification holds up even when the underlying AI returns a malformed response — a single hiccup no longer drops your tags.
- Memories that merge into an existing near-duplicate now get classified too.
🚀 Upgrading
No migration. No schema change. Zero downtime. Just deploy v1.9.0 — existing memories keep working exactly as before, and new memories pick up the new smarts automatically. (Backfilling your older memories with status/kind is coming soon.)