Lians vs Supermemory.
Supermemory is the universal memory API, scored on the same five invariants we score ourselves. Credit where it's strong, receipts where it isn't.
Credit first
Where Supermemory is genuinely strong.
Supermemory is built for speed and breadth: very fast ingestion across sources and connectors, hybrid retrieval with metadata filters, and a consolidated user-profile endpoint. As a drop-in “remember everything” layer for consumer AI apps, it delivers.
Regulated-memory eval · last run 2026-07-04
The five invariants.
| Regulated invariant | Lians | Supermemory |
|---|---|---|
| Stale revision suppressed | pass | partial |
| Point-in-time (as-of) recall | pass | absent |
| Provable erasure (crypto-shred + cert) | pass | partial |
| Lookahead / backtest guard | pass | absent |
| Audit-state snapshot at T | pass | absent |
| Score (pass = 1, partial = ½) | 5.0 / 5 | 1.0 / 5 |
Where the lanes diverge
Different jobs, different guarantees.
It is accumulate-and-rank by design: no supersession decision on write, no as-of or valid-time queries, no admission control in front of the store, and deletion without proof. Lians governs what enters memory, closes stale windows deterministically, and can hand an examiner the state at any past T.
The honest recommendation
Which should you pick?
Choose Supermemory for consumer apps that need broad, fast ingestion and retrieval. Choose Lians when what's in memory, and what was in memory, carries regulatory weight.
add/search map onto add/recall 1:1; moving means gaining event-time stamps, keyed supersession, and the audit chain without changing your agent loop.