#Content Addressing

The Data Hierarchy showed how a Mind is organized into a tree of Avatars, Experiences, Memories, and Engrams. Each node in that tree needs an address so the network can find it, store it, and serve it to anyone who asks. On today's internet, addresses are locations. On Avatarnet, addresses are fingerprints.

#Location Addressing

On the traditional web, an address tells you where a file lives.

"Get me the file at https://server.com/einstein/physics-papers/time-dilation.json"

This depends on a server being online. If the server goes down, the content is gone. If someone changes the file at that URL, you get different content at the same address with no way to tell. You must trust the server to give you the right thing.

#Content Addressing

On Avatarnet, an address tells you what the content is.

"Get me the content whose SHA-512 hash is a1b2c3d4e5f6..."

It does not matter who gives it to you. It does not matter where they are. You can verify it yourself: hash what you received and check whether the result matches the address. If it matches, you got the right content, guaranteed by mathematics. If it does not match, the data was tampered with or corrupted, and you throw it away.

The analogy: Location addressing is like saying "go to 123 Main Street." Content addressing is like saying "find me the book whose ISBN is 978-3-16-148410-0." You can get that book from any library, any bookstore, or any friend, and the ISBN proves it is the right book regardless of where it came from.

#How It Works

The mechanism is the SHA-512 hash function, the same algorithm the Signing and Encryption page explained under Mind Integrity. You feed in any input, from a single character to an entire book, and SHA-512 produces a fixed 64-byte fingerprint.

Content:  "Time dilation is one of the most striking predictions..."
                    │
                    ↓  SHA-512 hash function
                    │
Hash:     a1b2c3d4e5f6a7b8...  (64 bytes, always the same length)

Same content   →  always the same hash
Different content  →  always a different hash
Cannot reverse →  the hash does not reveal the content

That 64-byte hash is the content's permanent address on the network. The Key and Hash Sizes page showed these are the same 64 bytes that encode to 128 hex characters or roughly 87 Base58 characters. You have already seen them. This page explains what they are for.

#CIDs and Multihash

The raw SHA-512 hash gets wrapped in a short header called a multihash before it becomes a usable address. The header is two bytes: one identifying the hash algorithm (SHA-512) and one stating the length (64 bytes). The result is 66 bytes total.

Multihash = [algorithm code] + [digest length] + [64 bytes of SHA-512 hash]
          = 66 bytes total

This self-describing format means every address carries proof of which algorithm produced it. If SHA-512 is ever superseded in the distant future, new addresses can use a different algorithm code and old addresses still declare what they are. Nothing breaks.

The full address, multihash and all, is called a Content Identifier, or CID. When Avatarnet stores a reference to an engram, a memory, or an experience, it stores the CID. When a visitor fetches content, they ask for it by CID. When any machine proves it has the data, it shows the CID matches.

#Verification Across the Tree

Content addressing does not only protect individual engrams. Because the hierarchy is a Merkle DAG, parent nodes include the CIDs of their children. This means a single root hash verifies the entire tree.

Suppose a visitor downloads the root manifest for einstein.avtr and gets the root CID. That root CID was computed from the avatar object, which includes the experience CIDs. Each experience CID was computed from an experience object that includes memory CIDs. Each memory CID includes engram CIDs. If any engram anywhere in the tree has been altered, its CID will not match, which means its parent memory's CID will not match, which means the experience CID will not match, which means the root CID will not match.

Root CID ← depends on experience CIDs
  └─ Experience CID ← depends on memory CIDs
       └─ Memory CID ← depends on engram CIDs
            └─ Engram CID ← depends on the engram's content

Tamper with one engram:
  → Engram CID changes
    → Memory CID changes
      → Experience CID changes
        → Root CID changes

The visitor re-hashes the root and gets a different result.
Tampering detected.

A verifier does not need to re-hash every engram to detect a problem. They can start from the root and walk down, re-hashing each level and comparing it to the CID the parent claims. The first mismatch pinpoints exactly where the tampering occurred.

#Why This Matters

Content addressing gives Avatarnet four guarantees that location-based systems cannot provide.

Tamper evidence. Changing a single character in an engram changes its CID, which changes its parent's CID, all the way to the root. Tampering is not just detectable; it is mathematically self-evident.

Deduplication. Identical content produces identical CIDs. If two people write the same text, the network stores one copy. If a machine already has an engram, it does not need to download it again.

Location independence. The CID does not care where the content lives. Any machine on the network can serve it, and the requester verifies by re-hashing. No single server needs to stay online.

Permanent addresses. An engram that never changes keeps its CID forever. That address is valid for as long as anyone on the network holds a copy. There is no expiration, no renewal, and no central authority that assigns or revokes it.

Content addressing guarantees that nobody can tamper with your data without detection. But not all data should be visible to everyone. Some engrams are public, some are shared with selected people, and some are for nobody but you. The network needs to know the difference, and the next page Privacy Spheres explains how that works.