Browse documentation
Guarantees and limits
Scope of the language, execution, authority and evidence contracts.
Each guarantee depends on the runtime version, the command used and the stated conditions.
Release perspective
These docs describe the planned first release. Current builds may lack some capabilities. The runtime is not publicly distributed yet; the tutorials require source access.
Language and tooling
| Property | Scope and limit |
|---|---|
| Rust compatibility | Documented Rust meaning for the supported subset. Arbitrary Rust programs or Cargo dependencies are not automatically supported. |
| Safety | Types, ownership and defined semantics reduce specific bugs. They do not establish business correctness. |
| Zero-native-build loop | The installed runtime checks and executes supported edits. Runtime/native builds and internal lowering still occur. |
| Test caching | Explicit cache use can reuse unchanged case outcomes. Verification and executable generation remain fresh in this workflow. |
| Content identity | Identifies exact content. It does not prove benign behaviour or publisher authenticity. |
Execution and authority
| Property | Scope and limit |
|---|---|
| Determinism | Defined behaviour under the code, inputs and execution contract. External delivery needs separate rules. |
| Durability | State and progress persist under the storage/recovery contract. No universal exactly-once external delivery. |
| Actor isolation | State ownership and typed communication. Not an automatic OS sandbox or separate deployment. |
| Scoped authority | Grants and authenticated context constrain operations. A declaration or handle grants no permission. |
| Communication inspection | Declared, permitted and observed views with coverage and context. Not complete information-flow proof. |
Trusted runtime code, native hosts, credentials, storage and deployment remain in the security boundary. Completing accepted work and disclosing protected data can require different authority decisions.
Evidence
| Evidence | Establishes |
|---|---|
| Check | Acceptance under particular rules and inputs |
| Test | Observed result for selected cases and environment |
| Replay | Reconstruction under retained inputs and execution rules |
| Benchmark | Performance for a specified workload and configuration |
| Proof | A stated property under formal assumptions |
| Live observation | Behaviour in a particular deployment, time and coverage window |
One category does not substitute for another. Comparative security and performance claims need a defined threat model or workload and reproducible evidence.
Separate capabilities
Cross-node execution, managed cloud hosting, total-local-storage-loss recovery, arbitrary Rust compatibility and a complete graph query-and-rules language are outside the first-release assumption. Framework construction is supported scope; a Rails-like framework is not bundled.