yuechen-li-dev/oct

Oct

Oct is a scientific programming language and toolchain for reproducible research.

It is designed for the point where notebooks and scripts stop being enough: when an experiment needs tests, units, artifacts, packages, native binaries, and a distribution story. Oct's guiding principle is that the correct way should also be the easiest way.

What is Oct?

Oct is an early scientific programming language/toolchain for portable computation, reproducible research, and AI-assisted experimentation.

Oct is built on Go as its systems substrate. Oct programs compile through Go, build quickly, run as native binaries, and target the platforms Go targets. Existing Go libraries can be exposed to Oct through explicit Octxiliary wrappers, letting researchers keep a high-level scientific language without losing access to the Go ecosystem.

The language includes first-class scientific features that are already represented in the repository's contracts and libraries: SI units, xUnit-style testing, arrays/vectors/matrices, native Einstein tensor notation, Octomata flow/state machines, utility scoring, fallible functions, package sync, optional lock.octagon reproducibility, and explicit native wrapper builds.

Why Oct?

Oct is for research code that has outgrown throwaway scripts but still needs to stay close to the scientist's model of the problem.

• Reproducibility by default: tests, artifacts, package manifests, and optional lockfiles are part of the normal workflow.
• Scientific language surface: units, tensors, arrays, matrices, fallible functions, and experiment artifacts are language/toolchain concerns rather than notebook conventions.
• Native distribution path: the current implementation compiles through Go, so the compiled path can produce ordinary native binaries.
• Explicit integration: Octxiliary sidecars expose Go libraries through manifest-declared wrappers instead of hidden ambient bindings.
• Agent-friendly workflow: an LLM can creat