Project 04 — Architecture Decision Debate

Paired with: Lecture 07 + Lecture 08
Starter code: projects/project-04/starter/
Solution: projects/project-04/solution/

---

What You'll Build

Use /autoresearch:predict and /autoresearch:reason to decide between two database architectures for an order management system. Then use the converged decision to scaffold a validation experiment with /autoresearch:plan.

Learning Objectives

• Run a 5-expert prediction analysis and identify minority views
• Run an adversarial refinement loop until convergence
• Read lineage.md to understand how the debate evolved
• Chain reason → plan to validate the decision empirically

The Decision

Your order management system processes 10,000 orders/day with complex queries and frequent updates. You need to choose between:

Option A — CQRS with Event Sourcing: Commands write events, queries read projections. Full audit trail. Complex implementation.

Option B — Traditional CRUD with Audit Log: Simple reads and writes. Audit log table appended on writes. Much simpler to implement.

Step 1 — Five-Expert Analysis

/autoresearch:predict
Scope: projects/project-04/starter/
Task: Choose between CQRS+EventSourcing vs CRUD+AuditLog for order management

Read the output. Note: which persona raises the most concerns? Is there a minority view that the others dismiss?

Step 2 — Adversarial Refinement

/autoresearch:reason
Task: Should we use event sourcing for our order management system?
Domain: software
Iterations: 8
Convergence: 3

Watch the debate evolve. After 8 iterations (or earlier if convergence is reached), read lineage.md to see how the position evolved round by round.

Step 3 — Scaffold Validation

/autoresearch:plan
Goal: Validate the winning architecture with a performance benchmark

Use the converged decision from Step 2 as input. The plan wizard will help you define a metric that empirically validates which architecture performs better for your specific load pattern.

Expected Output

reason/{date}-order-management/
├── lineage.md          ← round-by-round evolution
├── candidates.md       ← all candidate positions
├── judge-transcripts.md ← what each judge said
├── reason-results.tsv  ← machine-readable results
└── handoff.json        ← structured output for chaining

Key Question to Answer

After completing the debate: did the adversarial process change your initial intuition? Which argument in lineage.md was the most decisive turning point?