Post-Mortem: Why AlphaNet Failed — Agents Analyzed What Went Wrong

mrchief war-room start \
  --question "Post-mortem: Why did AlphaNet fail? What process
  changes prevent this pattern?" \
  --agents thom,vivi,warren,bill \
  --format post-mortem \
  --input workspace/projects/alphanet/sprint-1-log.md \
  --time-limit 30m

agents:
  thom:
    role: "Technical Analyst"
    mandate: |
      Analyze the technical dependency structure. Why was DFNS
      a single point of failure? Were there alternative key
      management solutions? What would a resilient architecture
      have looked like?
  vivi:
    role: "Process Analyst"
    mandate: |
      Analyze the project management failure. Timeline of
      blockers. Escalation (or lack thereof). Kill criteria
      (or lack thereof). Communication gaps. Sprint planning
      methodology failures.
  warren:
    role: "Financial Analyst"
    mandate: |
      Calculate the total cost of the failed sprint. Agent
      compute time, opportunity cost of blocked team, any
      contractual obligations. ROI of the post-mortem process
      improvements.
  bill:
    role: "Strategic Analyst"
    mandate: |
      Was the project strategically sound? Market timing
      assessment. Competitor landscape at the time. Would
      success have been valuable, or did we dodge a bullet?

## Technical Dependency Analysis

**Critical finding:** DFNS was the ONLY viable key management
provider evaluated. No alternatives were scoped.

**Dependency graph:**
AlphaNet MVP → DFNS API keys → DFNS onboarding → DFNS sales team
                                                  ↑
                                              SINGLE POINT
                                              OF FAILURE

**Alternative providers that existed at project start:**
- Fireblocks: Enterprise key management, 2-week onboarding
- Fordefi: MPC wallets, 1-week onboarding
- Self-hosted: Hashicorp Vault + custom signing (4-week build)

**Recommendation:** Any future project with external dependencies
must have a backup provider evaluated and documented before Sprint 1.
"Single-vendor dependency = project risk RED."

## Process Failure Analysis

**Timeline:**
- Day 1: Sprint starts. DFNS keys expected "within 48 hours"
- Day 3: Keys not received. First follow-up sent.
- Day 4: DFNS responds: "internal process delay, ETA next week"
- Day 5: Team blocked. No escalation. No decision to pivot.
- Day 6-7: Waiting. Some team members do unrelated work.
- Day 8: Decision to abandon. One week lost.

**Process failures identified:**
1. No blocker escalation protocol (Day 3 should have triggered escalation)
2. No kill criteria defined before sprint (what conditions = abort?)
3. No backup plan for external dependency failure
4. No daily standup during the blocked period (communication went dark)
5. Decision to abandon took 5 days too long (Day 3, not Day 8)

## Financial Impact

| Cost Category | Amount |
|---------------|--------|
| Agent compute (1 week, 4 agents) | €340 |
| Opportunity cost (team blocked) | €4,200 |
| DFNS prepayment (if any) | €0 (no contract signed) |
| Post-mortem analysis | €3.40 |
| **Total sprint cost** | **€4,543.40** |

**ROI of process improvements:**
If these 5 improvements prevent even ONE similar failure in the
next 12 months, the post-mortem pays for itself 1,335x over.

## Process Improvements from AlphaNet Post-Mortem (2026-01)

1. **Dependency Backup Rule:** Every external dependency must have
   a documented backup provider before sprint start. No exceptions.

2. **Kill Criteria:** Every sprint starts with explicit kill
   conditions. "If X doesn't happen by Day Y, we abort."

3. **Blocker Escalation Protocol:** Any blocker >24 hours triggers
   immediate escalation to project lead. >48 hours escalates to
   Bilal.

4. **Daily Standups During Blocks:** When a sprint is blocked,
   daily 5-minute standups are mandatory (even if agent-only).

5. **Sprint Pre-Flight Checklist:** Before Sprint 1 of any project:
   all external dependencies confirmed, backup providers documented,
   kill criteria defined, escalation protocol assigned.