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.

HGM validation record · 2026-10-06

The update completes the research-console and bounded benchmark-analysis workflow. It does not certify a live machine-labor deployment.

Corrected behavior

Repeatable verification

tests/ covers the original engine/configuration/lineage/owner behavior plus queued-budget regression cases, constrained concurrent work, pending agent ceilings, horizon reservations, no-work owner pause, invalid configuration and undefined-ratio feedback.

web/tests/model.test.mjs checks the generated reference/constrained/paused records, exact arithmetic and source binding, forged approvals, cross-source candidates, invalid imports, the synthetic task contract, all ten work-order categories and market percentage arithmetic.

web/tests/browser-qa.mjs checks desktop/mobile widths (320, 390, 768, 1024 and 1440), accessibility, local diagrams, scenario changes, paused ratios, analysis download, accepted and rejected candidate checks, work-order download, invalid import state clearing, market inputs and the preserved console. It fails on page errors and external asset requests. Evidence is saved under reports/pages/hgm/ and uploaded by CI.

The HGM workflow runs all these checks plus the shared-worker task inspector. The Pages workflow verifies the integrated public route. Repository-wide gates remain required before merge.

What remains deployment-specific

No real OpenClaw or Work provider session, unrelated reviewer acceptance, buyer use, settlement, 72-hour host commissioning or proof of universal task competence is established by this change. The connected adapter is an available, fail-closed path that needs protected configuration and exact operator admission. Imported records and unsigned worker receipts do not authenticate their own provenance.

Simulated strategy performance is sensitive to seeds, different action rates, input distributions, cost assumptions and pending work. The original ROI field is a gross multiple. The $40 trillion opportunity is a user-supplied planning scenario; the page neither sources it as a measured fact nor treats it as revenue.

← Back to Huxley–Gödel Machine