You access Cubbit through a frontend application that lets you store, download, sync and share your files.
In the backend, your files are encrypted with AES-256, fragmented into 24 pieces, out of which 12 copies are made to ensure redundancy, for a total of 36 chunks.
Each chunk is then saved on a different Cubbit Cell. The location of each chunk is established by an AI coordinator that sits on Cubbit’s proprietary servers.
When you access a file remotely that you saved on Cubbit, you are actually asking the AI coordinator where your chunks are, after which you connect to the corresponding Cells via p2p end-to-end encrypted channels, download the chunks, reassemble the file and decrypt it with the key that is known to you and you alone.
The marvelous part is that the AI coordinator is cryptographically blind, since it never gets to know neither the content of the file nor the encryption key.