Value Stream Map
Where does time die when creating a venture on the platform?
Current State (Manual)
[NEED] → Wire Composition → Wire A2A → Scope DB → Register Agent → Verify → [VENTURE]
Root Endpoints Tables Card Boot
| Step | Cycle Time | Wait Time | %C&A | Bottleneck? |
|---|---|---|---|---|
| Wire composition root | 4h | 0h | 60% | No |
| Wire A2A endpoints | 4h | 2h (decision on namespace) | 50% | Yes — requires arch knowledge |
| Scope database tables | 8h | 4h (migration review) | 70% | Yes — 18 tables × manual FK |
| Register agent card | 2h | 1h (naming convention) | 80% | No |
| Verify boot | 4h | 0h | 30% | Yes — no test script existed |
Lead time: 29 hours (over 3-5 days with context switching) Cycle time: 22 hours Flow efficiency: 22 / 29 = 76% (misleadingly high — real efficiency is low because the entire stream runs once per venture)
Future State (Generator)
[NEED] → nx g sprout → Boot → Accept → [VENTURE]
(templates) (auto) (script)
| Step | Cycle Time | Wait Time | %C&A |
|---|---|---|---|
| Run generator | 2 min | 0 | 95% |
| Cold boot | 30 sec | 0 | 90% |
| Run acceptance test | 1 min | 0 | 99% |
Lead time: ~4 minutes Improvement: 29 hours → 4 minutes (435x reduction)
Seven Wastes Analysis
| Waste | Current State | Generator Eliminates? |
|---|---|---|
| Waiting | 7h decision/review waits | Yes — decisions encoded in templates |
| Rework | 30% boot failures from wiring errors | Yes — tested templates |
| Motion | Context switching across 18 tables | Yes — single command |
| Over-processing | Manual FK validation per table | Yes — generator knows schema |
| Transport | Copy-paste from drmg-sales patterns | Yes — additive templates |
Key Finding
The bottleneck isn't any single step — it's that the stream runs at all. The generator eliminates the entire manual stream, not just the slow parts.
Context
- Outcome Map — What success looks like
- Dependency Map — What blocks the generator from running
- Capability Map — Maturity of each component