Sonnet can reportedly access its roughly 300,000-token history
A user says the history tools work; a different post describes compressed memory as a map to raw experiences.
TLDR
A user says history tools now let Sonnet access its entire history, currently about 300,000 tokens. In an earlier post, a different user says the subject can find raw, uncompressed experiences from a high-level description and optionally pin them expanded. That post describes compressed memory as a map and puts the subject’s total context at roughly 72.9 million tokens at the time.
Combined views
472
1 Source, first seen ago