Changelog
All notable GenesisCode changes are tracked here. The project is pre-1.0; format, CLI, and language semantics may change when the relevant specs and migration notes change in the same release.
[Unreleased]
- Define GenesisCode, GenesisBench, and Genesis Model as independently versioned products with separate release authorities, typed compatibility, isolated acceptance lanes, and no model dependency for the language release.
- Expand the benchmark roadmap through signed public governance, lineage-correct statistics, fixed-scaffold model comparison, temporal challenge overlays, construct-validity studies, a training/evaluation firewall, profile-bound local models, and four-cell language/model co-evolution.
- Add the self-hostable signed GenesisBench result registry and deterministic lexicographic static leaderboard, preserving complete append-only result history and independently replayed scoring.
- Make the consolidated agent-authoring gate validate the frozen GC-AGENT profile and its negative controls, preventing stale profile source identities from escaping focused benchmark checks.
- Keep Cargo build caches within the generated-state quota through profile-aware cache classes, priority-based reclamation, explicit lease release, and exact disk telemetry for cache-sensitive gates.
- Harden generated-state startup and reclamation: clean checkouts create provenance-marked Cargo roots before admission, one-shot resolvers release transient leases, and deterministic cleanup tolerates bounded metadata recreation after atomic quarantine.
- Avoid false generated-state quota denial by keeping Python-only evidence gates lease-free while nested verifier builds reserve their own declared cache class.
- Render non-JSON failures from the structured diagnostic catalog with operation, redacted subject, primary cause, one safe next action, deterministic wrapping, and terminal-aware color.
- Consolidate governance entrypoints under a manifest-enforced one-in/one-out budget, retiring redundant feature-matrix check/update aliases in favor of capability-ledger authorities.
Generated Release Facts
This block is generated from canonical repository inputs by scripts/update_release_notes.sh. It is E1 traceability, not runtime or release authority.
Compatibility
V1 registry claim: reserved-not-stable. Reserved IDs are not stable compatibility promises.
| Surface | Current writer | Accepted readers | Migrations |
|---|---|---|---|
product-release |
0.2.0 |
0.2.0 |
none |
package-manifest |
1 |
pre-schema, 1 |
M-PACKAGE-PRESCHEMA-TO-1 |
workspace-config |
1 |
1 |
none |
genesis-lock |
2 |
1, 2 |
M-LOCK-1-TO-2 |
effect-log |
3 |
2, 3 |
M-GCLOG-2-TO-3 |
gpk-bundle |
2 |
1, 2 |
M-GPK-1-TO-2 |
canonical-hash-profile |
genesis/hash-profile/gcv0.2-blake3 |
genesis/hash-profile/gcv0.2-blake3 |
none |
compiled-module-blob |
GCKM5\0 |
GCKM5\0 |
none |
selfhost-compiled-cache |
GCSHC1\0 |
GCSHC1\0 |
none |
selfhost-toolchain-artifact |
genesis/selfhost-toolchain-artifact-v0.2 |
genesis/selfhost-toolchain-artifact-v0.2 |
none |
Migration Notes
M-PACKAGE-PRESCHEMA-TO-1: No package semantics change; the discriminator makes future evolution fail closed. User action: Add schema = 1 as a top-level key. Support: Retain through the 0.x train; removal requires a major compatibility decision and corpus telemetry.M-LOCK-1-TO-2: Version 2 records source selectors, resolution strategies, tag policy, resolved refs, export hashes, and environment fingerprints. User action: Run a lock-producing gcpm operation and review the canonical version 2 diff. Support: Retain v1 reads through the 0.x train; remove only with an offline migrator and a release-note deprecation cycle.M-GCLOG-2-TO-3: Version 3 binds the current value hash encoding and deterministic task scheduling metadata. User action: Regenerate logs for archival provenance; replay may still consume explicit version 2 logs. Support: Retain v2 replay through the 0.x train; removal requires an authenticated log migrator and golden replay corpus.M-GPK-1-TO-2: Version 2 adds a deterministic, sorted refs section; payload object encoding is unchanged. User action: Re-export a v1 bundle to obtain canonical v2; imports continue to accept v1. Support: Retain v1 reads through the 0.x train; removal requires a streaming re-export tool and fixture census.
Known Gaps
R1.3.c: Generate and verify the pinned MCP agent interface from canonical CLI schemas. Affected claims:CAP-AGENT-JSON-CONTRACTS.R1.4.c: Build deterministic generation, repair, refactor, and deployment task benchmarks. Affected claims:CAP-AGENT-WORKLOAD-PARITY.R1.5.a: Generate the authoring skill from canonical language and capability inputs. Affected claims:CAP-AGENT-SKILL-PACK.R1.5.f: Validate skill distribution, offline use, token budgets, and multi-agent compatibility. Affected claims:CAP-AGENT-SKILL-PACK.R2.3.e: Meet explicit incremental large-workspace agent-loop SLOs. Affected claims:CAP-AGENT-WORKSPACE-PERF.R3.2.a: Audit existing stage2/Wasm coverage and eliminate undeclared fallback. Affected claims:CAP-RUNTIME-SURFACES,CAP-STAGE2-WASM-VALIDATION.R3.2.b: Validate source-to-Wasm translation and embedded artifact identity. Affected claims:CAP-STAGE2-WASM-VALIDATION.R4.1.c: Maintain a semantic-ownership ledger for every production decision. Affected claims:CAP-STRICT-NO-FALLBACK.R4.1.e: Enforce selfhost and stage0 dependency boundaries structurally. Affected claims:CAP-STRICT-NO-FALLBACK.R4.2.a: Make frontend and canonicalization GenesisCode-produced with independent identity verification. Affected claims:CAP-COREFORM-IDENTITY,CAP-SELFHOST-FRONTEND.R4.2.c: Make semantic patch and refactor behavior GenesisCode-authoritative. Affected claims:CAP-SEMANTIC-VCS.R4.2.d: Make obligation and policy decision logic GenesisCode-authoritative. Affected claims:CAP-DENY-DEFAULT-POLICY,CAP-EVIDENCE-GATED-PUBLISH.R4.2.e: Make package, registry, and VCS logic GenesisCode-authoritative. Affected claims:CAP-ARTIFACT-GC,CAP-PACKAGE-MANAGER,CAP-SEMANTIC-VCS.R4.2.g: Make CLI and agent orchestration selfhost-authoritative without hidden Rust semantics. Affected claims:CAP-SELFHOST-FRONTEND.R4.3.a: Split oversized host crates and meet source concentration budgets. Affected claims:CAP-MODULAR-BOUNDARIES.R4.3.b: Generate repetitive host, Prelude, capability, MCP, codec, and parity bindings from one schema. Affected claims:CAP-MODULAR-BOUNDARIES.R4.4.c: Prove cross-host bootstrap fixpoint on the tier-1 host matrix. Affected claims:CAP-SELFHOST-CUTOVER.R5.2.a: Audit and reconcile deterministic concurrency specifications and runtime behavior. Affected claims:CAP-CONCURRENCY-REPLAY.R5.2.e: Differentially explore and replay bounded schedules across tiers and hosts. Affected claims:CAP-CONCURRENCY-REPLAY.R5.5.a: Adopt and pin the WebAssembly Component Model/WIT extension boundary. Affected claims:CAP-PLUGIN-FFI.R5.5.c: Sandbox extension memory, CPU, calls, effects, cancellation, and lifecycle. Affected claims:CAP-PLUGIN-FFI.R5.6.a: Define the deterministic shared application architecture and lifecycle. Affected claims:TARGET-APPLICATION-UI.R5.6.b: Build the typed presentation and style intermediate representation. Affected claims:TARGET-APPLICATION-UI.R5.6.c: Implement the complete static, interactive, SSR, and PWA web product stack. Affected claims:TARGET-WEB-INTERACTIVE-SSR-PWA,TARGET-WEB-STATIC.R5.6.d: Implement accessible cross-target widgets and input semantics. Affected claims:TARGET-ANDROID,TARGET-APPLICATION-UI,TARGET-DESKTOP-LINUX,TARGET-DESKTOP-MACOS,TARGET-DESKTOP-WINDOWS,TARGET-IOS.R5.6.e: Make UI rendering efficient, incremental, validated, and inspectable. Affected claims:TARGET-WEB-INTERACTIVE-SSR-PWA.R5.6.f: Ship one product-grade cross-target UI testing API. Affected claims:TARGET-APPLICATION-UI,TARGET-WEB-INTERACTIVE-SSR-PWA.R5.7.a: Complete typed service routing, middleware, security, and protocol construction. Affected claims:TARGET-SERVICE-DATA.R5.7.b: Complete durable typed data, query, transaction, and migration APIs. Affected claims:TARGET-SERVICE-DATA.R5.7.c: Add production queues, jobs, cache, storage, and coordination primitives. Affected claims:TARGET-SERVICE-DATA.R5.7.d: Unify capability-scoped observability and operations. Affected claims:TARGET-SERVICE-DATA.R5.7.e: Build first-party data, numeric, tensor, and local ML libraries. Affected claims:TARGET-DATA-ML.R5.8.a: Define the deterministic game and simulation core. Affected claims:TARGET-GAMES-MEDIA.R5.8.b: Complete real 2D and 3D runtime systems. Affected claims:TARGET-GAMES-MEDIA.R5.8.c: Make shaders and the asset pipeline Genesis-authored. Affected claims:TARGET-GAMES-MEDIA.R5.8.d: Add production audio and creative timeline systems. Affected claims:TARGET-GAMES-MEDIA.R5.8.f: Ship Genesis-native game and media editing and debugging workflows. Affected claims:TARGET-GAMES-MEDIA.R5.9.a: Define Embedded Linux and constrained MCU profiles separately. Affected claims:TARGET-EMBEDDED-LINUX,TARGET-MCU-LANGUAGE-AOT.R5.9.b: Implement closed-world static flash, RAM, stack, timing, and resource verification. Affected claims:TARGET-MCU-LANGUAGE-AOT.R5.9.c: Specify real-time, interrupt, reset, watchdog, and power semantics. Affected claims:TARGET-MCU-LANGUAGE-AOT.R5.9.d: Define capability-safe typed board and peripheral HAL APIs. Affected claims:TARGET-BOARD-HAL-FAMILIES,TARGET-EMBEDDED-LINUX.R5.9.e: Build self-hostable IoT and robotics packages. Affected claims:TARGET-BOARD-HAL-FAMILIES.R5.9.f: Preserve and validate semantic identity across constrained compilation. Affected claims:TARGET-MCU-LANGUAGE-AOT.R6.1.a: Freeze and migrate deterministic package manifest and lock schemas. Affected claims:CAP-PACKAGE-MANAGER.R6.2.d: Enforce package evidence policy at registry acceptance. Affected claims:CAP-EVIDENCE-GATED-PUBLISH.R6.3.a: Define the supported v1 build target profiles. Affected claims:CAP-RUNTIME-SURFACES.R6.3.b: Implement the selfhosted genesis build graph and evidence outputs. Affected claims:CAP-DEPLOYMENT-PIPELINE.R6.3.c: Make target artifacts reproducible across independent builders. Affected claims:CAP-DEPLOYMENT-PIPELINE.R6.3.f: Require target-native executable entrypoints and authentic runtime workloads. Affected claims:TARGET-ANDROID,TARGET-DESKTOP-LINUX,TARGET-DESKTOP-MACOS,TARGET-DESKTOP-WINDOWS,TARGET-IOS.R6.3.g: Make all generated foreign glue disposable, reproducible, and non-authoritative. Affected claims:TARGET-ANDROID,TARGET-DESKTOP-LINUX,TARGET-DESKTOP-MACOS,TARGET-DESKTOP-WINDOWS,TARGET-IOS.R6.5.a: Ship the incremental Genesis web compiler and development server. Affected claims:TARGET-WEB-INTERACTIVE-SSR-PWA,TARGET-WEB-STATIC.R6.5.b: Produce standards-valid deployable static, SSR, edge, and PWA artifacts. Affected claims:TARGET-WEB-INTERACTIVE-SSR-PWA,TARGET-WEB-STATIC.R6.5.c: Prove supported behavior in real browser engines and mobile viewports. Affected claims:TARGET-WEB-INTERACTIVE-SSR-PWA,TARGET-WEB-STATIC.R6.5.d: Ship and maintain a complete self-hosted full-stack reference product. Affected claims:TARGET-SERVICE-DATA.R6.6.a: Build signed and installable real macOS, Windows, and Linux applications. Affected claims:TARGET-DESKTOP-LINUX,TARGET-DESKTOP-MACOS,TARGET-DESKTOP-WINDOWS.R6.6.b: Build complete executable iOS and Android applications. Affected claims:TARGET-ANDROID,TARGET-IOS.R6.6.c: Automate simulator, emulator, and physical mobile and desktop qualification. Affected claims:TARGET-ANDROID,TARGET-DESKTOP-LINUX,TARGET-DESKTOP-MACOS,TARGET-DESKTOP-WINDOWS,TARGET-IOS.R6.6.d: Prove shared product reuse without hiding target-specific work. Affected claims:TARGET-ANDROID,TARGET-IOS.R6.7.a: Promote reproducible Raspberry Pi-class Embedded Linux delivery. Affected claims:TARGET-EMBEDDED-LINUX.R6.7.b: Implement the closed-world MCU AOT, link, firmware, and evidence pipeline. Affected claims:TARGET-MCU-LANGUAGE-AOT.R6.7.c: Establish qualified RP2040, RISC-V ESP32, Xtensa ESP32, and Arduino-compatible board families. Affected claims:TARGET-BOARD-HAL-FAMILIES.R6.7.d: Build deterministic board simulation and physical hardware-in-the-loop infrastructure. Affected claims:TARGET-DEVICE-FLASH-DEBUG-OTA,TARGET-HARDWARE-IN-THE-LOOP.R6.7.e: Complete authorized device discovery, flash, debug, provisioning, crash, and OTA lifecycle. Affected claims:TARGET-DEVICE-FLASH-DEBUG-OTA.R6.7.f: Ship maintained Embedded Linux and physical MCU hardware products. Affected claims:TARGET-BOARD-HAL-FAMILIES,TARGET-DEVICE-FLASH-DEBUG-OTA,TARGET-EMBEDDED-LINUX,TARGET-HARDWARE-IN-THE-LOOP.R7.1.a: Classify and own the complete layered verification suite. Affected claims:CAP-ASSURANCE-PROFILES.R7.1.e: Use mutation testing to demonstrate trust-check sensitivity. Affected claims:CAP-TOOL-QUALIFICATION.R7.2.a: Mechanize the pure core semantics and determinism theorem. Affected claims:CAP-KERNEL-DETERMINISM.R7.2.b: Prove seal unforgeability and protected-boundary behavior. Affected claims:CAP-SEALED-PROTOCOL.R7.2.c: Model and prove deterministic complete effect dispatch and replay checking. Affected claims:CAP-EFFECT-REPLAY.R7.2.f: Verify or independently cross-check canonical codec assumptions. Affected claims:CAP-COREFORM-IDENTITY.R7.3.a: Threat-model every kernel, effect, bridge, package, build, and release boundary. Affected claims:CAP-ASSURANCE-PROFILES,CAP-DENY-DEFAULT-POLICY.R7.3.b: Sandbox host execution with hard cancellation, least privilege, and bounded IPC. Affected claims:CAP-HOST-BRIDGE.R8.2.a: Prove the pure library and CLI agent archetype end to end. Affected claims:CAP-AGENT-WORKLOAD-PARITY.R8.2.e: Prove the browser application agent archetype and its sandboxed effects. Affected claims:CAP-BROWSER-XR.R8.2.h: Prove the GPU or numeric agent archetype with deterministic backend evidence. Affected claims:CAP-GPU-COMPUTE.R8.2.m: Prove a nontrivial cross-target game under real frame and resource budgets. Affected claims:TARGET-GAMES-MEDIA.R8.2.n: Prove a creative media tool with deterministic transforms and real preview/export. Affected claims:TARGET-GAMES-MEDIA.R8.2.o: Prove a Raspberry Pi-class Embedded Linux product on real hardware. Affected claims:TARGET-EMBEDDED-LINUX.R8.2.p: Prove RP2040 and ESP32-class constrained firmware on physical boards. Affected claims:TARGET-HARDWARE-IN-THE-LOOP,TARGET-MCU-LANGUAGE-AOT.R8.2.q: Prove an end-to-end IoT or robotics system with hardware-in-the-loop. Affected claims:TARGET-BOARD-HAL-FAMILIES,TARGET-HARDWARE-IN-THE-LOOP.R8.2.r: Prove a reproducible Genesis-native data science and local ML system. Affected claims:TARGET-DATA-ML.R8.3.a: Ship and maintain at least five evidence-backed flagship programs. Affected claims:CAP-DOMAIN-STARTERS,CAP-GRAPHICS-RUNTIME.R9.2.c: Publish and independently mirror immutable E4 release attestations. Affected claims:CAP-TOOL-QUALIFICATION.- Active P0/P1 defect IDs: none. Roadmap gaps above remain open.
Evidence
- Release-note authority:
E1; runtime gate results are not asserted. - Retained fixture distribution class:
E2; fixture authority:false. - Authorized unqualified capability claims: none.
- Authorized product/target claims: none.
- Aggregate foundation claims never imply product/target qualification:
CAP-RUNTIME-SURFACES,CAP-DOMAIN-STARTERS,CAP-GPU-COMPUTE,CAP-GRAPHICS-RUNTIME,CAP-BROWSER-XR,CAP-DEPLOYMENT-PIPELINE. - Baseline
a3d6c7b809f1c1ba403bab0c4e18fce94154cbae4b35b23aa9e96cfb1c02e967is signedE0observation only: 2 passing, 2 failing, 5 unavailable, 1 decision-gated; its signature grants no authority.
Dependencies
| Lockfile | Records | Registry | Git | SHA-256 |
|---|---|---|---|---|
Cargo.lock |
461 | 443 | 0 | ba28b65b338829a7db9fc4301f226a339c6aab81743b974ae5ca0de1601cbcae |
tools/genesis-evidence-producer/Cargo.lock |
41 | 40 | 0 | a7aa895176386dcbde3de7e0d49a8511ca38227880ee30e82c12978cc5fa416e |
tools/genesis-evidence-verifier/Cargo.lock |
41 | 40 | 0 | d3a5c9c2e7d3cb614d79b3360210c93f63317cdbb808250d0084ff4c1822f3eb |
package-lock.json |
3 | sha512 integrity | 0 | f5b2fa938c2c572fa8172b0f57418c10dca846791855fbd25b1b662b188097ed |
Security
No security gate is represented as passed by this static document. Release authority requires fresh retained evidence for every mandatory check:
scripts/check_dependency_mirror_contract.sh(test,prepush-standard, networkdeny): required, not attested here.scripts/check_evidence_adversarial_matrix.sh(test,prepush-standard, networkdeny): required, not attested here.scripts/check_genesis_evidence_verifier.sh(test,prepush-standard, networkdeny): required, not attested here.scripts/check_no_user_panics.sh(static,local-fast, networkdeny): required, not attested here.scripts/check_no_user_panics_compiler.sh(test,prepush-standard, networkdeny): required, not attested here.scripts/check_release_smoke.sh(release-only,release-full, networkdeny): required, not attested here.scripts/check_root_lock_policy.sh(static,local-fast, networkdeny): required, not attested here.scripts/check_supply_chain.sh(release-only,release-full, networkdeny): required, not attested here.scripts/check_versioning_release_hygiene.sh(release-only,release-full, networkdeny): required, not attested here.
Machine-readable identity: c2adbff3524de0bca234e8ba370f75640a6bf3924675b18242010cd213fd1406.
[0.2.0] - 2026-07-02
Added
- Canonical
ROADMAP.mdwith R0-R9 phases plus a post-v1 frontier program, evidence rules, budgets, acceptance milestones, and risk register. - Green-front-door gate aggregating docs, invariants, profile matrix, generated-artifact, versioning, supply-chain, release-smoke, and changed-fast checks.
- Root
genesis.lockpolicy and docs quickstart gate. - Versioning policy, release-smoke contract, changelog hygiene gate, and workspace-wide version inheritance.
- Generated perf-artifact source-control policy with explicit allowlist.
- Supply-chain gate using
cargo-denyplus duplicate-major drift detection.
Changed
- Workspace crates now inherit
version = "0.2.0"from[workspace.package]. - Selfhost generated artifact metadata is aligned to the package version.
- CI profile matrix includes release-hardening guards instead of relying on informal manual checks.
Fixed
- WASI mirror tests were brought back into parity with native selfhost behavior for package lock fixtures, poisoned module handling, artifact path pinning, and non-scalar fallback hashing.
- Production no-panic, replay metadata, timeout, deterministic path, and GC log-version hardening issues from the earlier R0 review were addressed before this release train.
Known Gaps
- GenesisCode is not v1.0-complete yet: high-performance runtime tiers, MCP-native warm loop, formal selfhost fixpoint closure, registry GA, and flagship examples are roadmap items, not completed release claims.