Repository guide · 0 diagrams

View source on GitHub ↗
Original documentation, preserved from the repository. Historical projections and scenario ambitions are not evidence of live performance. See the current readiness record for deployment requirements.

Validation scope

The release combines an executable six-phase evaluator, a complete public viewer, preserved mission simulations and the existing admitted computer-work adapter. All original scenarios, version directories, reports and systems maps are retained.

Required checks

Check Evidence established
Original Python suite and all twelve CLI rehearsals Preserved models run; fresh artifacts and nonzero failures are observable
TypeScript scenario regressions Required validator quorum, valid identities, bounded automation and finite projections
Six-phase tests Exact money, qualification, routing, review/budget admission, full phase coverage and deterministic replay
Project proposal exports All twelve scopes and acceptance criteria preserved in browser/CLI JSON and Markdown, source binding, separate commissioning and no execution authorization
Adversarial artifact tests Incorrect content is rejected even after a matching hash is recomputed
Real adapter with six loopback fixtures Exact task parsing, admitted job/deployment binding, replay prevention and returned artifact checks
Browser QA Keyboard access, mobile layout, automated accessibility, downloads, invalid-file handling, changed review scope, stale-import rejection, overlapping-operation recovery, zero-capacity behavior, failure states and every preserved diagram
Pages checks Public catalog integration, publication build and published revision

The independent content checker does not import the producer's calculations. Shared schema, canonicalization and task-contract utilities define the comparison boundary. A bug shared by the specification and both implementations remains possible; substantive acceptance still needs independent judgment.

Evidence boundaries

The reference data and worker observations are synthetic. Qualification thresholds, prices, review durations, useful-output assumptions and market figures are scenario inputs. Content hashes prove byte integrity, not authorship, truth or economic value.

No check here commissions a live provider, runs on an operator's physical desktop, establishes unrelated reviewer acceptance, proves demand, settles paid jobs or operates physical infrastructure. Those require environment-specific commissioning and recorded end-to-end evidence. No release claim should describe this synthetic rehearsal as completed customer work or a fully qualified unattended production deployment.

The preserved dashboards show configured simulation metrics, including historical “CI”, readiness and probability labels. Their visible fixture notice applies to every such metric. Actual release checks come from GitHub Actions for the exact commit, not those stored values.

Preservation

The original Mermaid blocks in every README remain verbatim. Original graph sources and versioned models remain available. The viewer quotes reserved punctuation for rendering and sanitizes SVG/HTML; it does not remove graph nodes to hide failures. Generated website assets use the locked Mermaid and DOMPurify dependencies and make no runtime CDN requests.

Historic radar syntax is mapped to the supported radar diagram without changing axes or values. Historic signed quadrant coordinates are mapped from [-1, 1] to Mermaid's [0, 1] display range; original coordinate pairs remain in every point label and the chart title states the original range. Group references and malformed edge delimiters are repaired only for display.

← Back to Meta-Agentic ALPHA AGI Jobs