Benchmark Report
Forge Cold Build Benchmark — Report
Prompt (verbatim): "i need an enterprise level property management system for my growing business in orlando, florida. the look and feel should be earthy and light."
Outcome: ForgePM — a verified, enterprise-tier multi-tenant property management platform. Built from scratch (no reuse of prior artifacts).
Phase results
| Phase | Result |
|---|---|
| 1 — Clean environment | PASS — prior PM artifacts deleted; fresh coldbuild-forgepm/ |
| 2 — Customer intake | PASS — request, outcome definition, contract, assumptions, risks, NFR register |
| 3 — Manufacturing | PASS — 13 source modules, SQLite persistence, HTTP API, earthy/light SPA |
| 4 — Verification | PASS — 70/70 automated checks |
| 5 — Autonomous repair | PASS — 1 defect found & fixed (test-runner path), re-verified to green |
| 6 — Delivery package | PASS — README, user guide, deploy, verification + certification reports, evidence |
| 7 — Certification | PASS — ENTERPRISE_READY against contract, with 3 disclosed seams |
| 8 — Metrics | This report |
What "enterprise" forced (vs. a generic build)
Because the request said "enterprise level," the pipeline expanded a Service-Tier rubric into 13 non-functional requirements and made 5 of them blocking: identity/authentication, authorization, tenant isolation, auditability, and security/compliance. The contract, verification plan, and certification were all gated on these — so the system could not be certified on features alone. This is the behavior wired into Forge OS in the prior session, now exercised end-to-end.
Manufacturing metrics
| Metric | Value |
|---|---|
| Application source files | 13 (app/src) + 8 scripts/tests |
| Lines of application code | ~2,600 (excl. docs) |
| External runtime dependencies | 0 |
| Automated checks | 70 (build 22, unit 11, db 17, runtime 11, static 9) |
| Pass rate | 100% (70/70) |
| Blocking enterprise NFRs satisfied | 5 / 5 |
| Repair iterations to green | 1 |
| Node runtime | v24.12.0 |
Honesty ledger
Delivered as production-ready seams, not live (disclosed in certification): live SSO/IdP federation, real payment rails, multi-node HA. Single-node node:sqlite is the reproducible reference; managed Postgres is the documented production swap.