Archive lineage
The readme identifies an untouched iCloud source tree, an original forensic ZIP and a working copy. The recovered MANIFEST_FULL.tsv supplies path, size, modification-time and hash records. The original full ZIP is a separate locator; its complete payload has not been materialized in this website session.
State artifacts recovered now
The website audit has actual bytes for the Seferim memory database and four Epona/Rhiannon state artifacts, as well as the readme and full manifest. The database contains event rows but no query-event or statistics rows in the named tables. The short Epona.state and Rhiannon.state files contain malformed JSON.
Fresh manifest calculation
The current read-only audit parsed all 122,596 manifest rows, with no duplicate paths or malformed size values. There are 55,125 distinct SHA-256 values, 1,274,723,818 bytes across all listed paths and 74,736,967 bytes when each distinct hash is counted once. These are reproducible index calculations. They do not establish that every listed payload was recovered in this website session; the full original forensic ZIP remains a separate source locator.
The stored learning text
The complete readable database export includes all five schemas, 35,338 learning events, the SQLite sequence record and an inventory of the 65 distinct item strings. Many events repeat introductory text about topics such as consciousness, natural language processing, music, game theory and automation. Some stored strings start with numeric fragments, so this archive also allows the ingestion fidelity to be examined directly. Stored text, event repetition and a confidence label do not establish that the content was validated or used successfully in a later response.