Published · The current foundation
v0.0.5-alpha.1 provides typed functions, contracts, numerical vectors and tiles, runtime budgets, and relative source modules. This is the published baseline, with experimental syntax and APIs.
Published alpha.6 · Source navigation foundations
v0.0.6-alpha.1 adds opt-in source symbol/reference inspection and a finite-vector rmsbuiltin. These capabilities remain experimental; there is no integrated LSP, incremental editor cache, GPU backend or native compilation command.
Published alpha.6.3 · Deterministic numerical primitives
The published alpha.6.3 update adds topk, topk_values,softmax_temp and logsoftmax_temp to the experimental interpreter. The previous alpha.6.2 release added optional Rust embedding caches, but not an automatic incremental CLI. Internal scalar IR validation does not mean that a general native compiler has shipped.
Next priorities · Reliable developer tooling
Source indexing offers a foundation for future editor integrations, but an actual language server, stable cross-edit identity and rename operations need separate implementation and testing. There is no guaranteed timetable.
Planned · Stronger language and library foundations
The source roadmap then calls for:
- Explicit constant-expression and library organization design.
- User-defined types and callable values.
- A compiler backend and developer tooling.
- Domain-specific libraries.
See the reviewed compiler roadmap for the upstream ordering. Priorities may change as implementation reveals new requirements.
Planned · Scientific computing and numerical AI
Scientific programming is a long-term direction, grounded in Scarp’s existing finite vectors, matrix kernels and numerical validation. Sampling-based integration and interval derivatives are being reviewed in compiler PR #153. These proposed functions are not in the published v0.0.6-alpha.3 compiler. General tensors, ODE solvers, automatic differentiation, GPU execution and stable scarp.*scientific namespaces are not implemented.
Read the scientific computing proposal for the release gates, architecture and feature dependencies. Proposed library names are not available imports.
Future vision · AI development with less friction
Scarp aims to make AI development more approachable through language and tooling design. Today’s numerical values are an early foundation. A complete AI framework, automatic differentiation, GPU execution, general tensors, and a mature package ecosystem require separate design and implementation milestones.
Toward v1.0.0
A stable release should be grounded in a documented language contract, reliable tools and packages, conformance tests, compatibility expectations, and clear limitations. These are readiness goals for this project hub, not a claim that the compiler team has committed to a fixed release checklist.
There is no announced v1.0.0 date. We track concrete capabilities and reviewed evidence instead of arbitrary completion percentages.