CAD3 Data
A universal encoding format for cryptographically verified, self-describing data. Every value is content-addressable, every structure is self-replicating.
Data is the foundation of everything in a decentralised network—global state, smart contracts, transactions, cryptographic proofs. If the encoding is wrong, nothing else matters.
Existing formats were never designed for this world. JSON has no canonical form. Protocol Buffers need external schemas. XML cannot build Merkle trees. None of them give data a stable, verifiable identity.
CAD3 was built from first principles to solve this: a single, universal encoding where every value has exactly one representation, every structure is cryptographically verifiable, and the entire lattice—from a single integer to petabytes of global state—shares one coherent format.
Beyond Legacy Formats
Why the world needed a new encoding
No Identity
JSON, XML, and most formats allow multiple encodings of the same value. Without a canonical form, there is no stable hash—and without a stable hash, there is no content addressing, no Merkle trees, no trustless verification.
No Integrity
Legacy formats treat cryptographic verification as someone else's problem. CAD3 makes it structural: every cell hashes to a Value ID, every tree is a Merkle DAG, and every piece of data carries its own proof of authenticity.
No Scale
Flat formats break when data grows beyond a single message. CAD3 encodes data as trees of bounded cells—a 4GB blob is just six levels deep. Partial synchronisation, streaming, and zero-copy operations come free.
Core Innovations
Eight design principles that make CAD3 unique
Canonical Encoding
Every value has exactly one binary representation. No ambiguity, no variation—one value, one encoding, one identity. This is the foundation that makes everything else possible.
Content-Addressable
Hash the encoding with SHA3-256 and you get a Value ID—a decentralised pointer that uniquely identifies any piece of data across the entire network, without coordination.
Self-Describing
Every encoded value carries its own type information. No external schemas, no version negotiation, no out-of-band metadata. Any participant can decode any value on sight.
Automatic Merkle DAGs
Large structures naturally form cryptographic hash trees. Every branch is independently verifiable, every subtree can be shared or validated without the whole. Integrity is structural, not bolted on.
Bounded Cells
No encoded cell exceeds 16KB. This fixed upper bound enables streaming, zero-copy operations, and predictable memory use—while trees of cells can represent data of unlimited size.
DoS Resistant
Encoding and decoding run in strict O(n) time and space. Invalid encodings are rejected immediately. Malicious payloads cannot trigger superlinear computation—security is guaranteed by the format itself.
Partial Data
Only transmit what changed. Large structures can be partially shared using cryptographic references to known subtrees—enabling efficient delta synchronisation across the network.
High Performance
Compact binary encoding, high branching factors, and structural sharing minimise both bandwidth and memory. A 4GB dataset is only six tree levels deep with less than 1% overhead.
Universal Data
One format for every kind of value
Primitives
Integers of arbitrary size, IEEE 754 doubles, booleans, nil, and Unicode characters — all with compact, canonical binary representations.
Text & Binary
UTF-8 strings, raw binary blobs, symbolic names, and keywords. Small values embed directly; large values form balanced trees that scale without limit.
Collections
Vectors, lists, hash maps, sets, and ordered indexes — all immutable, structurally shared, and content-addressable. O(log n) updates with automatic Merkle verification.
Cryptographic Values
First-class Ed25519 signed values and content-addressed references. Digital signatures and hash-linked pointers are native to the format, not an afterthought.
Smart Contracts & Code
CVM bytecode, transactions, and the entire global state are encoded in CAD3. Code is data. The same format that stores a number stores a program.
Application-Defined
Sparse records, coded values, and extension types let applications define their own semantics while remaining fully compatible with the CAD3 ecosystem.
Built for the Future
Properties that no other format delivers together
Infinite Scale
Balanced trees of bounded cells represent data of any size. Structural sharing means updates copy only the changed path. The lattice grows without bound while each participant handles only what they need.
Zero Ambiguity
One value, one encoding, one hash. Canonical representation eliminates an entire class of security vulnerabilities and makes every piece of data independently verifiable by anyone, anywhere.
Universal Compatibility
CAD3 naturally encodes JSON, XML, CSV, S-expressions, binary data, and content-addressable storage. It does not replace these formats—it subsumes them, giving each a canonical, verifiable representation.
Explore the stack
See how CAD3 underpins the lattice, or dive into the full encoding specification.