Tensor

tensor@1 is the Forge Pool primitive family for dense numeric workloads, matrix operations, tensor transforms, and computation-adjacent AI plumbing.

What tensor@1 is For

Typical uses include:

matrix multiplication
tensor transforms
embedding operations
vector and dense numeric routines
inference pre-processing and post-processing
distributed numeric reduction

Example Identity

json

{
  "op": {
    "name": "tensor",
    "version": 1,
    "profile": "matmul.v1"
  }
}

Relationship to Older MatMul Framing

Earlier documentation may describe matrix multiplication as a standalone route or numeric demo.

Inside the Kernel model, that capability is better understood as:

primitive family: tensor@1
profile: matmul.v1

This keeps the public system clean while allowing tensor execution to grow into a broader family over time.

Expected Output Surfaces

Tensor workloads may produce:

result tensors or matrices
shape metadata
numeric precision metadata
verification summaries
artifact references
replay identifiers

Why Tensor Belongs in the Kernel

Dense numeric execution is not separate from the rest of Forge Pool. It is one of the primitive computational classes that the system should support natively.

By placing tensor workloads inside the Kernel family model, Forge Pool keeps a unified execution doctrine across simulation, search, graph, media, and numeric compute.

Tensor ​

What tensor@1 is For ​

Example Identity ​

Relationship to Older MatMul Framing ​