Verifiable at gigabyte scale

Content addressing only works on big files if you can chunk, diff, and stream-verify them — git stores a fresh copy of a large binary on every edit. mkit cuts files at content-defined boundaries (FastCDC), records the chunk list in a ChunkedBlob, ships only the changed chunks as a delta, and verifies each chunk against the root hash as it arrives (Bao). Drop a file — or let the auto-editor run — and watch all four below.