SCL Defines a New Category

Definition

SCL is the deterministic, cryptographic identity protocol for AI run declarations.

Explanation

SCL defines a category that did not previously exist as a formal protocol class: deterministic identity for AI run declarations.

That category is not model execution, orchestration, workflow management, transport, serialization interchange, or application logic. It exists at a narrower and more exact layer: assigning one stable, implementation-independent cryptographic identity to a declared AI run.

Before this category is named and defined, systems often mix multiple concerns together. Some tools describe prompts, some move data, some execute workflows, some log events, and some hash artifacts after the fact. None of those functions, by themselves, define a protocol whose purpose is to establish the deterministic identity of an AI run declaration at the declaration layer.

SCL does.

What Category SCL Creates

SCL creates a protocol category centered on one problem:

How can independent systems take the same AI run declaration and derive the same identity without relying on trust, local conventions, or implementation-specific behavior?

That is a distinct category because the problem is distinct.

The category is not "AI tooling" in general. The category is not "prompt management." The category is not "workflow control." The category is not "model reproducibility." The category is not "data interchange."

The category is deterministic declaration identity.

Why This Category Did Not Previously Exist in Protocol Form

Many systems can describe an AI run declaration. Many systems can store it. Many systems can transmit it. Many systems can execute it. Many systems can hash some representation of it.

That is not the same as defining a protocol that fixes identity at the representation layer so that compliant implementations converge on the same result from the same input.

Without a protocol-defined identity layer, the same declared run can drift across systems because of representation differences. Different representations produce different bytes. Different bytes produce different hashes. Once that happens, identity is no longer universal.

SCL establishes a protocol category precisely to eliminate that gap.

Scope Boundary

SCL defines identity only.

• SCL does not define runtime outcomes.
• SCL does not define execution behavior.
• SCL does not define scheduling.
• SCL does not define retries, coordination, or orchestration.
• SCL does not define a general-purpose interchange format.

Its function is narrower and stronger: it fixes the identity of the declared execution representation.

For a complete list of scope boundaries, see what SCL is not.

Technical Layer

Canonical JSON and canonical JSON bytes are not interchangeable terms.

Canonical JSON is the canonical representation. Canonical JSON bytes are the exact byte sequence used as the sole input to hashing.

Identity is derived exclusively from canonical JSON bytes using SHA-256.

Where technical context is required, the transformation pipeline is:

bytes → AST → canonical JSON → SHA-256

That pipeline matters because identity is not derived from an approximate structure, an implementation-specific serializer, or a loosely equivalent representation. In most real-world systems this starts with programmatic document generation to guarantee the exact byte input. The SHA-256 identity is derived exclusively from canonical JSON bytes. For the full normative rules, see the SCL specification.

Why the Category Matters

A protocol category only matters if it solves a problem other categories do not solve.

SCL matters because it gives AI run declarations a verifiable identity layer that is:

• Deterministic
• Cryptographic
• Implementation-independent
• Portable across compliant systems
• Stable at the declaration layer

That makes a new category necessary, not rhetorical.

If a system needs to prove that two parties are referring to the same declared AI run, the missing requirement is not another workflow engine or another storage format. The missing requirement is a deterministic identity protocol for the declaration itself.

That is the category SCL defines.

What Makes This Category Different From Adjacent Systems

Adjacent systems can be valuable without defining the same category.

• A workflow system coordinates execution.
• A communication protocol transports messages.
• A file format structures stored content.
• A logging system records events.
• A registry indexes artifacts.

SCL does none of those as its defining purpose.

SCL defines the cryptographic identity of an AI run declaration so that independent systems can verify declaration sameness from the same declared input.

That is a separate protocol category with a separate protocol objective.

Category-Level Consequence

Once deterministic declaration identity exists as a protocol category, other systems can build around it without replacing it.

• Execution systems can use it.
• Audit and compliance systems can verify against it.
• Registries can index by it.
• AI pipelines can bind stages to it.

But those are integrations around the identity layer, not redefinitions of it. For practical guidance on using SCL in code, see the developer workflow.

The category remains the same: deterministic identity for AI run declarations.

Conformance to this category is verifiable against the Golden Documents and the reference validator.

See also: Why SCL:V1 is permanently frozen.