Factory apps validation — architecture decision
Goal: Decide, per candidate factory app (9 total), whether Robert should BUILD standalone, BUILD ON Zaruba’s services, ADOPT off-the-shelf, or DROP — given Zaruba’s rebuild-from-audit doctrine and today’s ground truth from groupon2 access.
Status: Draft — dossier complete, decision pending Robert’s acceptance (flip to active when accepted).
Trigger: SSH access to groupon2/monorepo-development + monorepo-infra landed 2026-07-18; Robert asked to re-validate every app decision via oh-architecture. Prior verdicts: research/00-dossier.md §2, analysis/legacy/SYNTHESIS.md §4.
Read order
| # | Doc | What it does |
|---|---|---|
| 00 | SYNTHESIS | 5-minute decision summary — matrix, diagrams, per-app verdicts, hardness re-score |
| 01 | Deep dive | Ground truth: Zaruba greenfield, infra, legacy estate — with file refs |
| 02 | Case studies | Industry evidence, agentic era + before |
| 03 | Options | 4 portfolio candidates (A Full Factory / B Zaruba-Native / C Two-Lane Thin / D Defer-Most) |
| 04 | Steelman | Advocate vs Prosecutor per option (8 independent agents) + cross-examination |
Out of scope
- Domain migration decisions themselves (Tomas Zaruba’s lane — Orders/Deal Factory scope, DB strategy, etc.)
- Infra bring-up (blocked on human-only GCP billing gate —
monorepo-infra) - The assessment deliverable for Tomas (separate:
research/11-monorepo-development-assessment-plan.md) - Old-estate (
groupon/groupon-monorepo) 1→N Encore split — separate program, unaffected here
Cross-references
| Path | What |
|---|---|
research/00-dossier.md |
Prior build-vs-buy verdicts (superseded in part by this plan) |
research/12-collab-split-and-fleet-monitor.md |
Two-lead collision protocol + fleet-monitor adopt/build verdict (incorporated) |
analysis/legacy/SYNTHESIS.md |
Legacy sample fleet analysis (evidence input) |
repos/monorepo-development/ |
Zaruba greenfield (system of record) |
repos/monorepo-infra/ |
Infra skeleton |