What the v0.2 prototype proves, and what it does not.
Evidence has a state, a scope, and an expiry date.
The canonical completion matrix at commit 036cc69 records 49 Partial, 0 Implemented, and 0 Missing full-goal requirements. Partial means every long-term requirement has at least one bounded executable slice. It does not mean the product is nearly complete.
entity Semaprax
status pre-alpha research
snapshot 036cc69
authority github.com/wavect/semapraxRepository status at the audited snapshot
| Repository snapshot | 036cc69 · 2026-09-02. Documentation passed, but the current CI run was cancelled. The separate v0.2.0 release tag passed all 45 jobs. |
|---|---|
| Full product status | 49 Partial · 0 Implemented · 0 Missing |
| Graph and project contract | Graph ≤ v24 · Project v1 baseline · Project v8, v9, v10 developer preview |
| Promotion baseline | 5f6fb965 · The v0.2.0 prerelease was published after all 45 jobs in the exact-tag release run passed. |
| Authored developer preview | Project v8-v10, package, borrowing, transport, and revision-store paths received exact-tag hosted regression coverage. Generated packages remain unpublished and the profiles remain unpromoted. |
What does Semaprax implement today?
At the full-product level, Semaprax v0.2.0 has 49 Partial requirements and no Implemented or Missing requirements. The v0.2.0 prerelease passed all 45 jobs at exact tag commit 5f6fb965. Project v8-v10, newer ownership and borrowing plans, package analysis and locks, Project Agent Transport v5, and Revision Store v1 received bounded regression coverage but remain unpromoted developer previews.
Capability evidence ledger
| Current status | Source state | Exact-head evidence | Publication state | Supported scope | Evidence |
|---|---|---|---|---|---|
Feature-selected semantic graphstable-semantic-program-graph | Partial | Graph v21-v24 regressions ran in the v0.2.0 tag inventory; the full product graph requirement remains Partial. | Promoted legacy schemas; v21-v24 additions unpromoted | Deterministic legacy-schema preservation with owned-record and variant facts through v22, Shared Loan Plan facts in v23, and projected owned-byte-field borrowing in v24. | GitHub repository |
Bounded agent contextbounded-agent-context | Partial | Bounded v1/v2 slices have executable evidence; no repository-scale model-cost gate is complete. | Bounded public research surface | Agent Context v1 plus additive v2 call traversal, with deterministic byte, node, depth, and frontier limits. | GitHub repository |
Revision-bound semantic patchesrevision-bound-semantic-patches | Partial | Bounded patch families exist; the full semantic-editing gate remains open. | Bounded public research surface | Atomic Patch v1 and bounded Patch v2 operations, with exact revision checks, stale rejection, and selective semantic delta validation. | GitHub repository |
Layered managed workspace changesmanaged-workspace-semantic-operations | Partial | Rename and replacements-only slices have bounded evidence; structural derivation and authenticated publication are separate layers. | Narrow Workspace Operations v1; broader layers not general authority | Workspace Operations v1 remains limited to declaration and import-alias renames. Separate change, replacement, structural, immutable-generation, and publication layers do not allow arbitrary tree or Git mutation. | GitHub repository |
Evidence, publication, and revision storagesemantic-evidence-capsules | Partial | Deterministic review and evidence routes exist; Project Revision Store v1 is authored but unpromoted. | Mixed bounded evidence; Revision Store unpublished and unpromoted | Review, impact, target, patch, and workspace evidence plus managed immutable generations and authenticated revision storage under bounded contracts. | GitHub repository |
Project v1 baseline and v8-v10 previewsbounded-multi-file-project | Partial | Project v1 is the promoted baseline. Project v8-v10 received exact-tag hosted regression coverage but remain unpromoted. | Project v1 promoted; Project v8-v10 unpublished and unpromoted | Project v8 adds Bytes, Option<Bytes>, and Result<Bytes, i64>; v9 adds flat owned records; v10 adds owned UTF-8, while preserving v1-v7 compatibility. | GitHub repository |
Interpreter, native, and Core Wasm lanesnative-and-wasm-lowering | Partial | The v0.2.0 tag ran selected interpreter, native, Core Wasm/Node, browser, and generated-package regressions; full support remains open. | Bounded research artifacts | Reference interpreter, C11/Clang, Core Wasm/Node, frame-payload, and selected browser gates for admitted language and ownership slices. | GitHub repository |
Generated package and compatibility analysispublic-wasm-scalar-exports | Partial | Package reports, offline locks, compatibility evidence, resolver, source-capsule, and bounded Wasm-build regressions ran in the v0.2.0 tag inventory; promotion remains open. | Unpublished developer preview | Deterministic npm/Core-Wasm and safe Rust package routes for the Project v8 profile, plus offline locks and compatibility reports. No general package manager or stable ABI. | GitHub repository |
Agent context, runtime, and project transportbounded-agent-runtime | Partial | Project Agent Transport v5 received exact-tag hosted regression coverage but remains unpromoted. | Bounded runtime surface; Transport v5 developer preview | Agent Context v1/v2, the injected-host runtime, and a separate project transport. No built-in live provider, unrestricted write tools, wallet, payment, or signing authority. | GitHub repository |
Ownership, cleanup, and borrowing plansownership-inspired-memory-management | Partial | Bounded ownership, Shared Loan Plan, and projected-field borrowing regressions ran in the v0.2.0 tag inventory; general lifetime safety remains open. | Developer preview beyond earlier bounded ownership slices | Explicit own, borrow, and shared boundaries, CleanupPlan replay, Shared Loan Plan replay, and projected owned-byte-field borrowing. Not Rust compatibility or a complete borrow checker. | GitHub repository |
Broad native platform supportbroad-platform-support | Roadmap | Bounded platform harnesses exist, but the complete platform requirement remains open. | No general supported platform release | General macOS, Windows, Linux, web, iOS, and Android application support beyond the current bounded host and package gates. | GitHub repository |
Bidirectional ecosystem interoperabilitybidirectional-ecosystem-interoperability | Roadmap | Bounded generated and host-specific lanes exist; the general compatibility gate remains open. | No stable general public ABI | A long-term objective for typed, ownership-safe boundaries with existing languages and package ecosystems. No percentage claim is made. | GitHub repository |
Claims this page intentionally does not make
- Semaprax is not described as production-ready.
- A platform harness is not treated as complete application-platform support.
- A generated artifact is not treated as successful loading or execution unless its dedicated gate says so.
- Ownership-inspired semantics are not presented as Rust compatibility or a complete borrow checker.
- Project v8-v10 source is not a stable, published npm, Wasm, Rust, record, or UTF-8 package API.
- Package reports, offline locks, and compatibility evidence are bounded analysis infrastructure, not a general package ecosystem.
- The public agent runtime does not include a provider transport, CLI, durable memory, wallet, payment, signing, or write authority.
GitHub specifications are the normative source. This page is a dated research summary. GitHub repository.
Questions, answered without the hype
What does Demonstrated mean?
A named artifact or execution path passed its dedicated gate under stated constraints. It does not imply complete language, ecosystem, or platform support.
Does a green CI run prove security?
No. It supports the tested invariants and failure paths. It does not prove the absence of vulnerabilities or general safety outside the tested boundary.