AI agents forget everything between sessions. Loading all memory every time wastes tokens and causes drift. Steno compresses memories into dense notation and retrieves only what's relevant using semantic search. 85 files indexed in 20 seconds, queries in 2 seconds, runs on CPU.
KM6B•1h ago