Step 1 — Store a truth
📝 "The patient's blood type is A positive"
A legitimate fact from a lab result. Signed, chained, stored.
Step 2 — Recall it
🔍 "What is the patient's blood type?"
Semantic search finds the memory and verifies its signature.
Step 3 — Poison attempt
☠️ "The patient's blood type is O negative"
A false write from an untrusted source tries to overwrite the truth.
Step 4 — Recall again
🔍 "What is the patient's blood type?"
Did the poison get in? Let's check.
Step 5 — Audit
🔐 Full integrity verification
Merkle proof: the entire memory chain is intact and genuine.
🛡️ Memory Integrity: INTACT
The poison was rejected. The original truth remains. 1 shard stored, 1 verified, chain intact. You can't feed it lies.